diff --git a/README.md b/README.md
index a383d98..d50a433 100644
--- a/README.md
+++ b/README.md
@@ -470,7 +470,7 @@ Update an existing work package.
 - `status_id` (integer, optional): Status ID
 - `priority_id` (integer, optional): Priority ID
 - `assignee_id` (integer, optional): User ID to assign to
-- `percentage_done` (integer, optional): Completion percentage (0-100)
+- `percentage_done` (integer, optional): Completion percentage (0-100). **Known instance behavior:** on some OpenProject instances progress is derived from status and `percentageDone` is read-only — sending it returns HTTP 422 `PropertyIsReadOnly`. The tool detects this, skips `percentage_done`, and still applies the other fields (with a note in the response).
 
 #### 13. `delete_work_package`
 Delete a work package.
diff --git a/pyproject.toml b/pyproject.toml
index f426c81..f265f11 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -20,7 +20,7 @@ classifiers = [
 ]
 
 dependencies = [
-    "fastmcp>=2.0.0",
+    "fastmcp>=2.0.0,<3",  # 2.x tool API (@mcp.tool -> FunctionTool.fn); 3.x changes this
     "mcp>=1.0.0",
     "aiohttp>=3.8.0",
     "python-dotenv>=1.0.0",
@@ -33,6 +33,7 @@ dependencies = [
 [project.optional-dependencies]
 dev = [
     "pytest>=7.0.0",
+    "pytest-asyncio>=0.21.0",
     "black>=22.0.0",
     "flake8>=4.0.0",
 ]
@@ -51,3 +52,7 @@ target-version = ['py38']
 [tool.flake8]
 max-line-length = 88
 extend-ignore = ["E203", "W503"]
+
+[tool.pytest.ini_options]
+asyncio_mode = "auto"
+testpaths = ["tests"]
diff --git a/src/tools/memberships.py b/src/tools/memberships.py
index a3d6b2d..330a8d9 100644
--- a/src/tools/memberships.py
+++ b/src/tools/memberships.py
@@ -1,12 +1,13 @@
 """Membership management tools."""
 
 from typing import Optional, List
-from pydantic import BaseModel, Field
+from pydantic import Field
 from src.server import mcp, get_client
+from src.utils.inputs import CoercibleModel
 from src.utils.formatting import format_success, format_error
 
 
-class CreateMembershipInput(BaseModel):
+class CreateMembershipInput(CoercibleModel):
     """Input model for creating memberships."""
     project_id: int = Field(..., description="Project ID", gt=0)
     user_id: Optional[int] = Field(None, description="User ID (required if group_id not provided)", gt=0)
@@ -16,7 +17,7 @@ class CreateMembershipInput(BaseModel):
     notification_message: Optional[str] = Field(None, description="Optional notification message")
 
 
-class UpdateMembershipInput(BaseModel):
+class UpdateMembershipInput(CoercibleModel):
     """Input model for updating memberships."""
     membership_id: int = Field(..., description="Membership ID to update", gt=0)
     role_ids: Optional[List[int]] = Field(None, description="New list of role IDs")
diff --git a/src/tools/news.py b/src/tools/news.py
index 7d2705a..8ac3de4 100644
--- a/src/tools/news.py
+++ b/src/tools/news.py
@@ -2,9 +2,10 @@
 
 import json
 from typing import Optional
-from pydantic import BaseModel, Field
+from pydantic import Field
 
 from src.server import mcp, get_client
+from src.utils.inputs import CoercibleModel
 from src.utils.formatting import (
     format_news_list,
     format_news_detail,
@@ -18,7 +19,7 @@
 # ============================================================
 
 
-class CreateNewsInput(BaseModel):
+class CreateNewsInput(CoercibleModel):
     """Input model for creating news."""
 
     project_id: int = Field(..., description="Project ID", gt=0)
@@ -29,7 +30,7 @@ class CreateNewsInput(BaseModel):
     description: str = Field(..., description="Main content (supports Markdown)")
 
 
-class UpdateNewsInput(BaseModel):
+class UpdateNewsInput(CoercibleModel):
     """Input model for updating news."""
 
     news_id: int = Field(..., description="News ID to update", gt=0)
diff --git a/src/tools/projects.py b/src/tools/projects.py
index 465529f..bf935f0 100644
--- a/src/tools/projects.py
+++ b/src/tools/projects.py
@@ -3,7 +3,8 @@
 import json
 from typing import Optional
 from src.server import mcp, get_client
-from pydantic import BaseModel, Field
+from pydantic import Field
+from src.utils.inputs import CoercibleModel
 from src.utils.formatting import format_success, format_error
 from src.utils.formatting import format_project_list
 
@@ -140,7 +141,7 @@ async def get_project(project_id: int) -> str:
         return f"❌ Failed to get project: {str(e)}"
 
 
-class CreateProjectInput(BaseModel):
+class CreateProjectInput(CoercibleModel):
     """Input model for creating projects."""
     name: str = Field(..., description="Project name", min_length=1, max_length=255)
     identifier: str = Field(..., description="Project identifier (lowercase, no spaces)", min_length=1, max_length=100)
@@ -150,7 +151,7 @@ class CreateProjectInput(BaseModel):
     parent_id: Optional[int] = Field(None, description="Parent project ID for sub-projects", gt=0)
 
 
-class AddSubprojectInput(BaseModel):
+class AddSubprojectInput(CoercibleModel):
     """Input model for adding subprojects."""
     parent_id: int = Field(..., description="Parent project ID", gt=0)
     name: str = Field(..., description="Subproject name", min_length=1, max_length=255)
@@ -159,7 +160,7 @@ class AddSubprojectInput(BaseModel):
     public: Optional[bool] = Field(None, description="Whether subproject is public")
 
 
-class UpdateProjectInput(BaseModel):
+class UpdateProjectInput(CoercibleModel):
     """Input model for updating projects."""
     project_id: int = Field(..., description="Project ID to update", gt=0)
     name: Optional[str] = Field(None, description="New project name", min_length=1, max_length=255)
diff --git a/src/tools/relations.py b/src/tools/relations.py
index 092b22a..68a49f4 100644
--- a/src/tools/relations.py
+++ b/src/tools/relations.py
@@ -1,12 +1,13 @@
 """Work package relation management tools (follows, blocks, relates, etc.)."""
 
 from typing import Optional
-from pydantic import BaseModel, Field
+from pydantic import Field
 from src.server import mcp, get_client
+from src.utils.inputs import CoercibleModel
 from src.utils.formatting import format_success, format_error
 
 
-class CreateRelationInput(BaseModel):
+class CreateRelationInput(CoercibleModel):
     """Input model for creating work package relations."""
     from_id: int = Field(..., description="Source work package ID", gt=0)
     to_id: int = Field(..., description="Target work package ID", gt=0)
@@ -15,7 +16,7 @@ class CreateRelationInput(BaseModel):
     description: Optional[str] = Field(None, description="Relation description")
 
 
-class UpdateRelationInput(BaseModel):
+class UpdateRelationInput(CoercibleModel):
     """Input model for updating work package relations."""
     relation_id: int = Field(..., description="Relation ID to update", gt=0)
     lag: Optional[int] = Field(None, description="New lag in working days")
diff --git a/src/tools/time_entries.py b/src/tools/time_entries.py
index c19a579..c994e8c 100644
--- a/src/tools/time_entries.py
+++ b/src/tools/time_entries.py
@@ -1,12 +1,13 @@
 """Time entry management tools for time tracking."""
 
 from typing import Optional
-from pydantic import BaseModel, Field
+from pydantic import Field
 from src.server import mcp, get_client
+from src.utils.inputs import CoercibleModel
 from src.utils.formatting import format_success, format_error
 
 
-class CreateTimeEntryInput(BaseModel):
+class CreateTimeEntryInput(CoercibleModel):
     """Input model for creating time entries."""
     work_package_id: int = Field(..., description="Work package ID", gt=0)
     hours: float = Field(..., description="Hours spent", gt=0)
@@ -15,7 +16,7 @@ class CreateTimeEntryInput(BaseModel):
     comment: Optional[str] = Field(None, description="Optional comment")
 
 
-class UpdateTimeEntryInput(BaseModel):
+class UpdateTimeEntryInput(CoercibleModel):
     """Input model for updating time entries."""
     time_entry_id: int = Field(..., description="Time entry ID to update", gt=0)
     hours: Optional[float] = Field(None, description="New hours spent", gt=0)
diff --git a/src/tools/versions.py b/src/tools/versions.py
index e417fbd..0f4fbff 100644
--- a/src/tools/versions.py
+++ b/src/tools/versions.py
@@ -1,12 +1,13 @@
 """Version/milestone management tools."""
 
 from typing import Optional
-from pydantic import BaseModel, Field
+from pydantic import Field
 from src.server import mcp, get_client
+from src.utils.inputs import CoercibleModel
 from src.utils.formatting import format_success, format_error
 
 
-class CreateVersionInput(BaseModel):
+class CreateVersionInput(CoercibleModel):
     """Input model for creating versions."""
     project_id: int = Field(..., description="Project ID", gt=0)
     name: str = Field(..., description="Version name", min_length=1, max_length=255)
diff --git a/src/tools/weekly_reports.py b/src/tools/weekly_reports.py
index 70af318..64b788f 100644
--- a/src/tools/weekly_reports.py
+++ b/src/tools/weekly_reports.py
@@ -3,9 +3,10 @@
 import json
 from typing import Optional
 from datetime import datetime, timedelta
-from pydantic import BaseModel, Field
+from pydantic import Field
 
 from src.server import mcp, get_client
+from src.utils.inputs import CoercibleModel
 from src.utils.formatting import format_success, format_error
 from src.utils.report_formatter import (
     format_weekly_report_markdown,
@@ -13,7 +14,7 @@
 )
 
 
-class GenerateWeeklyReportInput(BaseModel):
+class GenerateWeeklyReportInput(CoercibleModel):
     """Input model for generating weekly reports."""
     project_id: int = Field(..., description="Project ID to generate report for", gt=0)
     from_date: str = Field(..., description="Report start date (YYYY-MM-DD)")
@@ -23,7 +24,7 @@ class GenerateWeeklyReportInput(BaseModel):
     format: str = Field("markdown", description="Output format: 'markdown' or 'json'")
 
 
-class GetReportDataInput(BaseModel):
+class GetReportDataInput(CoercibleModel):
     """Input model for getting raw report data."""
     project_id: int = Field(..., description="Project ID", gt=0)
     from_date: str = Field(..., description="Start date (YYYY-MM-DD)")
diff --git a/src/tools/work_packages.py b/src/tools/work_packages.py
index 5348d08..ab777af 100644
--- a/src/tools/work_packages.py
+++ b/src/tools/work_packages.py
@@ -1,8 +1,8 @@
 """Work package management tools - Priority CRITICAL tools for 12 users."""
 
 import json
-from typing import Optional
-from pydantic import BaseModel, Field
+from typing import Annotated, Optional
+from pydantic import Field
 
 from src.server import mcp, get_client
 from src.utils.formatting import (
@@ -13,36 +13,11 @@
 )
 
 
-# Pydantic models for type-safe input validation
-class CreateWorkPackageInput(BaseModel):
-    """Input model for creating work packages with validation."""
-
-    project_id: int = Field(..., description="Project ID where work package will be created", gt=0)
-    subject: str = Field(..., description="Work package title/subject", min_length=1, max_length=255)
-    type_id: int = Field(..., description="Type ID (use list_types to see available types)", gt=0)
-    description: Optional[str] = Field(None, description="Detailed description in raw format")
-    start_date: Optional[str] = Field(None, description="Start date in ISO format (YYYY-MM-DD)")
-    due_date: Optional[str] = Field(None, description="Due date in ISO format (YYYY-MM-DD)")
-    assignee_id: Optional[int] = Field(None, description="Assignee user ID", gt=0)
-    status_id: Optional[int] = Field(None, description="Status ID", gt=0)
-    priority_id: Optional[int] = Field(None, description="Priority ID", gt=0)
-    version_id: Optional[int] = Field(None, description="Version/milestone ID to assign work package to", gt=0)
-
-
-class UpdateWorkPackageInput(BaseModel):
-    """Input model for updating work packages with validation."""
-
-    work_package_id: int = Field(..., description="Work package ID to update", gt=0)
-    subject: Optional[str] = Field(None, description="New subject/title", min_length=1, max_length=255)
-    description: Optional[str] = Field(None, description="New description")
-    type_id: Optional[int] = Field(None, description="New type ID", gt=0)
-    status_id: Optional[int] = Field(None, description="New status ID", gt=0)
-    priority_id: Optional[int] = Field(None, description="New priority ID", gt=0)
-    assignee_id: Optional[int] = Field(None, description="New assignee user ID", gt=0)
-    start_date: Optional[str] = Field(None, description="New start date (YYYY-MM-DD)")
-    due_date: Optional[str] = Field(None, description="New due date (YYYY-MM-DD)")
-    percentage_done: Optional[int] = Field(None, description="Progress percentage (0-100)", ge=0, le=100)
-    version_id: Optional[int] = Field(None, description="Version/milestone ID to assign work package to", gt=0)
+# create_work_package / update_work_package use explicit (flat) parameters rather
+# than a single wrapped Pydantic model. This produces a self-describing object
+# schema and avoids the opaque-parameter encoding that some MCP clients serialize
+# as a JSON string. The other tools keep wrapped models with the CoercibleModel
+# base as a safety net (see src/utils/inputs.py).
 
 
 @mcp.tool
@@ -53,37 +28,32 @@ async def list_work_packages(
     active_only: bool = True,
     offset: int = 0,
     page_size: int = 20,
-    
     # NEW: Multi-value filters (comma-separated IDs)
     priority_ids: Optional[str] = None,
     type_ids: Optional[str] = None,
     status_ids: Optional[str] = None,
     version_ids: Optional[str] = None,
-    
     # NEW: Date filters
     due_before: Optional[str] = None,  # YYYY-MM-DD
-    due_after: Optional[str] = None,   # YYYY-MM-DD
+    due_after: Optional[str] = None,  # YYYY-MM-DD
     created_after: Optional[str] = None,  # YYYY-MM-DD
     updated_after: Optional[str] = None,  # YYYY-MM-DD
-    
     # NEW: Boolean filters
     unassigned_only: bool = False,
     overdue_only: bool = False,
-    
     # NEW: Percentage filters
     percentage_done_min: Optional[int] = None,
     percentage_done_max: Optional[int] = None,
-    
     # NEW: Additional filters
     author_id: Optional[int] = None,
     parent_id: Optional[int] = None,
-    no_parent_only: bool = False
+    no_parent_only: bool = False,
 ) -> str:
     """List work packages (tasks) with advanced filtering - CRITICAL tool for flexible task search.
-    
+
     This is the most powerful search tool with 20+ filter parameters for finding exactly
     the tasks you need. Supports multiple filters combined with AND logic.
-    
+
     Args:
         # Basic filters
         project_id: Optional project ID to filter by
@@ -91,35 +61,35 @@ async def list_work_packages(
         active_only: If True, only show open work packages (default: True)
         offset: Starting index for pagination (default: 0)
         page_size: Number of results per page (default: 20, max: 100)
-        
+
         # Multi-value filters (comma-separated IDs)
         priority_ids: Comma-separated priority IDs (e.g., "3,4" for high+urgent)
         type_ids: Comma-separated type IDs (e.g., "1,2" for bugs+features)
         status_ids: Comma-separated status IDs (overrides active_only if provided)
         version_ids: Comma-separated version/sprint IDs
-        
+
         # Date filters
         due_before: Due date before this date (YYYY-MM-DD format)
         due_after: Due date after this date (YYYY-MM-DD format)
         created_after: Created after this date (YYYY-MM-DD format)
         updated_after: Updated after this date (YYYY-MM-DD format)
-        
+
         # Boolean filters
         unassigned_only: If True, only show tasks without assignee
         overdue_only: If True, only show tasks past their due date
-        
+
         # Percentage filters
         percentage_done_min: Minimum completion percentage (0-100)
         percentage_done_max: Maximum completion percentage (0-100)
-        
+
         # Additional filters
         author_id: Filter by task creator/author
         parent_id: Filter by parent work package ID (child tasks)
         no_parent_only: If True, only show top-level tasks (no parent)
-    
+
     Returns:
         Formatted list of work packages matching all specified filters
-        
+
     Examples:
         Find high-priority bugs due this week:
         {
@@ -128,14 +98,14 @@ async def list_work_packages(
             "due_before": "2025-12-15",
             "due_after": "2025-12-08"
         }
-        
+
         Find overdue unassigned tasks in project #5:
         {
             "project_id": 5,
             "unassigned_only": true,
             "overdue_only": true
         }
-        
+
         Find nearly complete tasks (>80%):
         {
             "percentage_done_min": 80,
@@ -144,19 +114,21 @@ async def list_work_packages(
     """
     try:
         from datetime import date, datetime
-        
+
         client = get_client()
 
         # Build filters list
         filters_list = []
-        
+
         # === STATUS FILTER ===
         # Priority: status_ids > overdue_only > active_only
         if status_ids:
             # Explicit status IDs provided
             status_list = [s.strip() for s in status_ids.split(",") if s.strip()]
             if status_list:
-                filters_list.append({"status": {"operator": "=", "values": status_list}})
+                filters_list.append(
+                    {"status": {"operator": "=", "values": status_list}}
+                )
         elif overdue_only:
             # Overdue mode: must be open
             filters_list.append({"status": {"operator": "o", "values": []}})
@@ -166,91 +138,163 @@ async def list_work_packages(
         else:
             # All statuses (open + closed)
             filters_list.append({"status": {"operator": "*", "values": []}})
-        
+
         # === ASSIGNEE FILTER ===
         if unassigned_only:
             # Unassigned takes priority over assignee_id
             filters_list.append({"assignee": {"operator": "!*", "values": []}})
         elif assignee_id:
-            filters_list.append({"assignee": {"operator": "=", "values": [str(assignee_id)]}})
-        
+            filters_list.append(
+                {"assignee": {"operator": "=", "values": [str(assignee_id)]}}
+            )
+
         # === PRIORITY FILTER ===
         if priority_ids:
             priority_list = [p.strip() for p in priority_ids.split(",") if p.strip()]
             if priority_list:
-                filters_list.append({"priority": {"operator": "=", "values": priority_list}})
-        
+                filters_list.append(
+                    {"priority": {"operator": "=", "values": priority_list}}
+                )
+
         # === TYPE FILTER ===
         if type_ids:
             type_list = [t.strip() for t in type_ids.split(",") if t.strip()]
             if type_list:
                 filters_list.append({"type": {"operator": "=", "values": type_list}})
-        
+
         # === VERSION FILTER ===
         if version_ids:
             version_list = [v.strip() for v in version_ids.split(",") if v.strip()]
             if version_list:
-                filters_list.append({"version": {"operator": "=", "values": version_list}})
-        
+                filters_list.append(
+                    {"version": {"operator": "=", "values": version_list}}
+                )
+
         # === DATE FILTERS ===
         # Overdue filter (special case)
         if overdue_only:
             # Due date < today
             today = date.today().isoformat()
-            filters_list.append({"dueDate": {"operator": "<>d", "values": ["2000-01-01", today]}})
+            filters_list.append(
+                {"dueDate": {"operator": "<>d", "values": ["2000-01-01", today]}}
+            )
         else:
             # Regular due date filters
             if due_before and due_after:
                 # Date range
-                filters_list.append({"dueDate": {"operator": "<>d", "values": [due_after, due_before]}})
+                filters_list.append(
+                    {"dueDate": {"operator": "<>d", "values": [due_after, due_before]}}
+                )
             elif due_before:
                 # Before specific date (use range from old date to due_before)
-                filters_list.append({"dueDate": {"operator": "<>d", "values": ["2000-01-01", due_before]}})
+                filters_list.append(
+                    {
+                        "dueDate": {
+                            "operator": "<>d",
+                            "values": ["2000-01-01", due_before],
+                        }
+                    }
+                )
             elif due_after:
                 # After specific date (use range from due_after to far future)
-                filters_list.append({"dueDate": {"operator": "<>d", "values": [due_after, "2099-12-31"]}})
-        
+                filters_list.append(
+                    {
+                        "dueDate": {
+                            "operator": "<>d",
+                            "values": [due_after, "2099-12-31"],
+                        }
+                    }
+                )
+
         # Created after filter
         if created_after:
             # Use date range from created_after to far future
-            filters_list.append({"createdAt": {"operator": "<>d", "values": [created_after, "2099-12-31"]}})
-        
+            filters_list.append(
+                {
+                    "createdAt": {
+                        "operator": "<>d",
+                        "values": [created_after, "2099-12-31"],
+                    }
+                }
+            )
+
         # Updated after filter
         if updated_after:
             # Use date range from updated_after to far future
-            filters_list.append({"updatedAt": {"operator": "<>d", "values": [updated_after, "2099-12-31"]}})
-        
+            filters_list.append(
+                {
+                    "updatedAt": {
+                        "operator": "<>d",
+                        "values": [updated_after, "2099-12-31"],
+                    }
+                }
+            )
+
         # === PERCENTAGE FILTER ===
         if percentage_done_min is not None and percentage_done_max is not None:
             # Range filter
             if percentage_done_min > percentage_done_max:
-                return format_error("percentage_done_min cannot be greater than percentage_done_max")
+                return format_error(
+                    "percentage_done_min cannot be greater than percentage_done_max"
+                )
             # Use two filters: >= min AND <= max
-            filters_list.append({"percentageDone": {"operator": ">=", "values": [str(percentage_done_min)]}})
-            filters_list.append({"percentageDone": {"operator": "<=", "values": [str(percentage_done_max)]}})
+            filters_list.append(
+                {
+                    "percentageDone": {
+                        "operator": ">=",
+                        "values": [str(percentage_done_min)],
+                    }
+                }
+            )
+            filters_list.append(
+                {
+                    "percentageDone": {
+                        "operator": "<=",
+                        "values": [str(percentage_done_max)],
+                    }
+                }
+            )
         elif percentage_done_min is not None:
             # Minimum only
             if percentage_done_min < 0 or percentage_done_min > 100:
                 return format_error("percentage_done_min must be between 0 and 100")
-            filters_list.append({"percentageDone": {"operator": ">=", "values": [str(percentage_done_min)]}})
+            filters_list.append(
+                {
+                    "percentageDone": {
+                        "operator": ">=",
+                        "values": [str(percentage_done_min)],
+                    }
+                }
+            )
         elif percentage_done_max is not None:
             # Maximum only
             if percentage_done_max < 0 or percentage_done_max > 100:
                 return format_error("percentage_done_max must be between 0 and 100")
-            filters_list.append({"percentageDone": {"operator": "<=", "values": [str(percentage_done_max)]}})
-        
+            filters_list.append(
+                {
+                    "percentageDone": {
+                        "operator": "<=",
+                        "values": [str(percentage_done_max)],
+                    }
+                }
+            )
+
         # === AUTHOR FILTER ===
         if author_id:
-            filters_list.append({"author": {"operator": "=", "values": [str(author_id)]}})
-        
+            filters_list.append(
+                {"author": {"operator": "=", "values": [str(author_id)]}}
+            )
+
         # === PARENT FILTER ===
         if no_parent_only:
             # Top-level tasks only (no parent)
             filters_list.append({"parent": {"operator": "!*", "values": []}})
         elif parent_id:
             # Specific parent
-            filters_list.append({"parent": {"operator": "=", "values": [str(parent_id)]}})
-        
+            filters_list.append(
+                {"parent": {"operator": "=", "values": [str(parent_id)]}}
+            )
+
         # Convert filters to JSON
         filters = json.dumps(filters_list) if filters_list else None
 
@@ -261,10 +305,7 @@ async def list_work_packages(
             return format_error("page_size must be between 1 and 100")
 
         result = await client.get_work_packages(
-            project_id=project_id,
-            filters=filters,
-            offset=offset,
-            page_size=page_size
+            project_id=project_id, filters=filters, offset=offset, page_size=page_size
         )
 
         work_packages = result.get("_embedded", {}).get("elements", [])
@@ -284,15 +325,13 @@ async def list_work_packages(
         return format_error(f"Failed to list work packages: {str(e)}")
 
 
-
-
 @mcp.tool
 async def search_work_packages(
     query: str,
     project_id: Optional[int] = None,
     active_only: bool = True,
     offset: int = 0,
-    page_size: int = 20
+    page_size: int = 20,
 ) -> str:
     """Search work packages by subject or ID - Fast search without pagination.
 
@@ -331,12 +370,9 @@ async def search_work_packages(
         filters_list = []
 
         # Add subjectOrId filter for search
-        filters_list.append({
-            "subjectOrId": {
-                "operator": "**",
-                "values": [query.strip()]
-            }
-        })
+        filters_list.append(
+            {"subjectOrId": {"operator": "**", "values": [query.strip()]}}
+        )
 
         # Add active_only filter if requested (same fix as list_work_packages)
         if active_only:
@@ -354,10 +390,7 @@ async def search_work_packages(
             return format_error("page_size must be between 1 and 100")
 
         result = await client.get_work_packages(
-            project_id=project_id,
-            filters=filters,
-            offset=offset,
-            page_size=page_size
+            project_id=project_id, filters=filters, offset=offset, page_size=page_size
         )
 
         work_packages = result.get("_embedded", {}).get("elements", [])
@@ -387,54 +420,83 @@ async def search_work_packages(
 
 
 @mcp.tool
-async def create_work_package(input: CreateWorkPackageInput) -> str:
+async def create_work_package(
+    project_id: Annotated[
+        int, Field(gt=0, description="Project ID where work package will be created")
+    ],
+    subject: Annotated[
+        str,
+        Field(min_length=1, max_length=255, description="Work package title/subject"),
+    ],
+    type_id: Annotated[
+        int, Field(gt=0, description="Type ID (use list_types to see available types)")
+    ],
+    description: Annotated[
+        Optional[str], Field(description="Detailed description in raw format")
+    ] = None,
+    start_date: Annotated[
+        Optional[str], Field(description="Start date in ISO format (YYYY-MM-DD)")
+    ] = None,
+    due_date: Annotated[
+        Optional[str], Field(description="Due date in ISO format (YYYY-MM-DD)")
+    ] = None,
+    assignee_id: Annotated[
+        Optional[int], Field(gt=0, description="Assignee user ID")
+    ] = None,
+    status_id: Annotated[Optional[int], Field(gt=0, description="Status ID")] = None,
+    priority_id: Annotated[
+        Optional[int], Field(gt=0, description="Priority ID")
+    ] = None,
+    version_id: Annotated[
+        Optional[int],
+        Field(gt=0, description="Version/milestone ID to assign work package to"),
+    ] = None,
+) -> str:
     """Create a new work package (task) - CRITICAL tool for creating tasks.
 
     This is one of the most important tools for your 12 users to create new work items.
 
     Args:
-        input: Work package data including project_id, subject, type_id, and optional fields
+        project_id: Project where the work package will be created.
+        subject: Work package title/subject.
+        type_id: Type ID (use list_types to see available types).
+        description: Detailed description in raw format.
+        start_date: Start date (YYYY-MM-DD).
+        due_date: Due date (YYYY-MM-DD).
+        assignee_id: Assignee user ID.
+        status_id: Status ID (accepted for compatibility; status on creation follows
+            the project/type default -- use update_work_package to change it).
+        priority_id: Priority ID.
+        version_id: Version/milestone ID.
 
     Returns:
         Success message with created work package ID and details
-
-    Example:
-        To create a bug in project 5:
-        {
-            "project_id": 5,
-            "subject": "Fix login issue",
-            "type_id": 1,
-            "description": "Users cannot login with valid credentials",
-            "priority_id": 3,
-            "assignee_id": 7,
-            "due_date": "2025-01-15"
-        }
     """
     try:
         client = get_client()
 
         # Build data dict for API
         data = {
-            "project": input.project_id,
-            "subject": input.subject,
-            "type": input.type_id,
+            "project": project_id,
+            "subject": subject,
+            "type": type_id,
         }
 
         # Add optional fields
-        if input.description:
-            data["description"] = input.description
-        if input.priority_id:
-            data["priority_id"] = input.priority_id
-        if input.assignee_id:
-            data["assignee_id"] = input.assignee_id
-        if input.version_id:
-            data["version_id"] = input.version_id
+        if description:
+            data["description"] = description
+        if priority_id:
+            data["priority_id"] = priority_id
+        if assignee_id:
+            data["assignee_id"] = assignee_id
+        if version_id:
+            data["version_id"] = version_id
 
         # Add date fields (use camelCase for API)
-        if input.start_date:
-            data["startDate"] = input.start_date
-        if input.due_date:
-            data["dueDate"] = input.due_date
+        if start_date:
+            data["startDate"] = start_date
+        if due_date:
+            data["dueDate"] = due_date
 
         # Create work package
         result = await client.create_work_package(data)
@@ -457,9 +519,9 @@ async def create_work_package(input: CreateWorkPackageInput) -> str:
         if "assignee" in embedded:
             text += f"**Assignee**: {embedded['assignee'].get('name', 'Unassigned')}\n"
 
-        if result.get('startDate'):
+        if result.get("startDate"):
             text += f"**Start Date**: {result['startDate']}\n"
-        if result.get('dueDate'):
+        if result.get("dueDate"):
             text += f"**Due Date**: {result['dueDate']}\n"
 
         return text
@@ -469,27 +531,58 @@ async def create_work_package(input: CreateWorkPackageInput) -> str:
 
 
 @mcp.tool
-async def update_work_package(input: UpdateWorkPackageInput) -> str:
+async def update_work_package(
+    work_package_id: Annotated[
+        int, Field(gt=0, description="Work package ID to update")
+    ],
+    subject: Annotated[
+        Optional[str],
+        Field(min_length=1, max_length=255, description="New subject/title"),
+    ] = None,
+    description: Annotated[Optional[str], Field(description="New description")] = None,
+    type_id: Annotated[Optional[int], Field(gt=0, description="New type ID")] = None,
+    status_id: Annotated[
+        Optional[int], Field(gt=0, description="New status ID")
+    ] = None,
+    priority_id: Annotated[
+        Optional[int], Field(gt=0, description="New priority ID")
+    ] = None,
+    assignee_id: Annotated[
+        Optional[int], Field(gt=0, description="New assignee user ID")
+    ] = None,
+    start_date: Annotated[
+        Optional[str], Field(description="New start date (YYYY-MM-DD)")
+    ] = None,
+    due_date: Annotated[
+        Optional[str], Field(description="New due date (YYYY-MM-DD)")
+    ] = None,
+    percentage_done: Annotated[
+        Optional[int], Field(ge=0, le=100, description="Progress percentage (0-100)")
+    ] = None,
+    version_id: Annotated[
+        Optional[int],
+        Field(gt=0, description="Version/milestone ID to assign work package to"),
+    ] = None,
+) -> str:
     """Update an existing work package (task) - CRITICAL tool for updating tasks.
 
     This is one of the most important tools for your 12 users to update work items,
-    including changing status, assignee, dates, and progress.
+    including changing status, assignee, dates, and progress. Only the fields you
+    pass are changed.
 
     Args:
-        input: Work package update data including work_package_id and fields to update
+        work_package_id: Work package to update.
+        subject, description, type_id, status_id, priority_id, assignee_id,
+        start_date, due_date, percentage_done, version_id: Optional fields; only
+        provided values are applied.
+
+    Note:
+        percentage_done is read-only on some OpenProject instances (progress is
+        derived from status/work). If the instance rejects it, that field is
+        skipped and the remaining fields are still applied.
 
     Returns:
         Success message with updated work package details
-
-    Example:
-        To update work package #123 status and assignee:
-        {
-            "work_package_id": 123,
-            "status_id": 5,
-            "assignee_id": 7,
-            "percentage_done": 50,
-            "due_date": "2025-01-20"
-        }
     """
     try:
         client = get_client()
@@ -497,34 +590,68 @@ async def update_work_package(input: UpdateWorkPackageInput) -> str:
         # Build data dict for API (only include provided fields)
         data = {}
 
-        if input.subject is not None:
-            data["subject"] = input.subject
-        if input.description is not None:
-            data["description"] = input.description
-        if input.type_id is not None:
-            data["type_id"] = input.type_id
-        if input.status_id is not None:
-            data["status_id"] = input.status_id
-        if input.priority_id is not None:
-            data["priority_id"] = input.priority_id
-        if input.assignee_id is not None:
-            data["assignee_id"] = input.assignee_id
-        if input.percentage_done is not None:
-            data["percentage_done"] = input.percentage_done
-        if input.version_id is not None:
-            data["version_id"] = input.version_id
+        if subject is not None:
+            data["subject"] = subject
+        if description is not None:
+            data["description"] = description
+        if type_id is not None:
+            data["type_id"] = type_id
+        if status_id is not None:
+            data["status_id"] = status_id
+        if priority_id is not None:
+            data["priority_id"] = priority_id
+        if assignee_id is not None:
+            data["assignee_id"] = assignee_id
+        if percentage_done is not None:
+            data["percentage_done"] = percentage_done
+        if version_id is not None:
+            data["version_id"] = version_id
 
         # Add date fields (use camelCase for API)
-        if input.start_date is not None:
-            data["startDate"] = input.start_date
-        if input.due_date is not None:
-            data["dueDate"] = input.due_date
+        if start_date is not None:
+            data["startDate"] = start_date
+        if due_date is not None:
+            data["dueDate"] = due_date
 
         if not data:
             return format_error("No fields provided to update")
 
-        # Update work package
-        result = await client.update_work_package(input.work_package_id, data)
+        # Update work package. percentage_done (percentageDone) is read-only on some
+        # OpenProject instances (progress is derived from status). If the API rejects
+        # it as read-only, drop that field and retry once so the rest of the update
+        # still applies rather than failing the whole request.
+        readonly_note = ""
+        try:
+            result = await client.update_work_package(work_package_id, data)
+        except Exception as exc:
+            # Detect via the stable v3 error identifier ("PropertyIsReadOnly") and
+            # API attribute name ("percentageDone"), both present in the 422 body,
+            # rather than the localizable human message -- so this keeps working if
+            # OpenProject rewords or translates the error text.
+            low = str(exc).lower()
+            pct_read_only = (
+                percentage_done is not None
+                and "percentage" in low
+                and (
+                    "propertyisreadonly" in low
+                    or "readonly" in low.replace("-", "").replace(" ", "")
+                    or "not writable" in low
+                )
+            )
+            if not pct_read_only:
+                raise
+            data.pop("percentage_done", None)
+            if not data:
+                return format_error(
+                    "percentage_done is read-only on this OpenProject instance "
+                    "(progress is derived from status), and no other fields were "
+                    "provided to update."
+                )
+            result = await client.update_work_package(work_package_id, data)
+            readonly_note = (
+                "\n⚠️ percentage_done is read-only on this instance and was "
+                "skipped; the other fields were updated."
+            )
 
         # Format success response
         wp_id = result.get("id")
@@ -544,13 +671,14 @@ async def update_work_package(input: UpdateWorkPackageInput) -> str:
         if "assignee" in embedded:
             text += f"**Assignee**: {embedded['assignee'].get('name', 'Unassigned')}\n"
 
-        if result.get('startDate'):
+        if result.get("startDate"):
             text += f"**Start Date**: {result['startDate']}\n"
-        if result.get('dueDate'):
+        if result.get("dueDate"):
             text += f"**Due Date**: {result['dueDate']}\n"
-        if 'percentageDone' in result:
+        if "percentageDone" in result:
             text += f"**Progress**: {result['percentageDone']}%\n"
 
+        text += readonly_note
         return text
 
     except Exception as e:
@@ -573,7 +701,9 @@ async def delete_work_package(work_package_id: int) -> str:
         success = await client.delete_work_package(work_package_id)
 
         if success:
-            return format_success(f"Work package #{work_package_id} deleted successfully")
+            return format_success(
+                f"Work package #{work_package_id} deleted successfully"
+            )
         else:
             return format_error(f"Failed to delete work package #{work_package_id}")
 
@@ -715,7 +845,7 @@ async def assign_work_package(work_package_id: int, assignee_id: int) -> str:
         if "assignee" in embedded:
             assignee_name = embedded["assignee"].get("name", "Unknown")
             text += f"**Assigned to**: {assignee_name}\n"
-        
+
         if "type" in embedded:
             text += f"**Type**: {embedded['type'].get('name', 'Unknown')}\n"
         if "status" in embedded:
@@ -723,7 +853,7 @@ async def assign_work_package(work_package_id: int, assignee_id: int) -> str:
         if "priority" in embedded:
             text += f"**Priority**: {embedded['priority'].get('name', 'Unknown')}\n"
 
-        if result.get('dueDate'):
+        if result.get("dueDate"):
             text += f"**Due Date**: {result['dueDate']}\n"
 
         return text
@@ -749,7 +879,9 @@ async def unassign_work_package(work_package_id: int) -> str:
 
         # Update work package with null assignee (unassign)
         # Note: We need to use the API directly since setting to None might not work
-        result = await client.update_work_package(work_package_id, {"assignee_id": None})
+        result = await client.update_work_package(
+            work_package_id, {"assignee_id": None}
+        )
 
         wp_id = result.get("id")
         wp_subject = result.get("subject")
@@ -772,9 +904,7 @@ async def unassign_work_package(work_package_id: int) -> str:
 
 @mcp.tool
 async def add_work_package_comment(
-    work_package_id: int,
-    comment: str,
-    internal: bool = False
+    work_package_id: int, comment: str, internal: bool = False
 ) -> str:
     """Add a comment/activity to a work package - CRITICAL for reporting and communication.
 
@@ -801,9 +931,7 @@ async def add_work_package_comment(
         client = get_client()
 
         result = await client.add_work_package_comment(
-            work_package_id=work_package_id,
-            comment=comment,
-            internal=internal
+            work_package_id=work_package_id, comment=comment, internal=internal
         )
 
         activity_id = result.get("id", "N/A")
@@ -811,7 +939,9 @@ async def add_work_package_comment(
         comment_html = comment_data.get("html", "")
         comment_raw = comment_data.get("raw", comment)
 
-        text = format_success(f"Comment added to work package #{work_package_id} successfully!\n\n")
+        text = format_success(
+            f"Comment added to work package #{work_package_id} successfully!\n\n"
+        )
         text += f"**Activity ID**: {activity_id}\n"
         text += f"**Internal**: {'Yes' if internal else 'No'}\n"
         text += f"**Comment**: {comment_raw[:200]}{'...' if len(comment_raw) > 200 else ''}\n"
@@ -853,7 +983,9 @@ async def list_work_package_activities(work_package_id: int) -> str:
         if not activities:
             return f"No activities found for work package #{work_package_id}."
 
-        text = format_success(f"Work Package #{work_package_id} Activities ({len(activities)}):\n\n")
+        text = format_success(
+            f"Work Package #{work_package_id} Activities ({len(activities)}):\n\n"
+        )
 
         for activity in activities:
             activity_id = activity.get("id", "N/A")
@@ -903,29 +1035,30 @@ async def list_work_package_activities(work_package_id: int) -> str:
 # ADVANCED FILTERS - New high-priority tools for better task discovery
 # ============================================================================
 
+
 @mcp.tool
 async def list_overdue_work_packages(
     project_id: Optional[int] = None,
     assignee_id: Optional[int] = None,
     priority_ids: Optional[str] = None,  # Comma-separated IDs like "3,4"
     type_ids: Optional[str] = None,  # Comma-separated IDs like "1,2"
-    page_size: int = 50
+    page_size: int = 50,
 ) -> str:
     """List all overdue work packages (tasks past their due date).
-    
+
     This tool helps identify tasks that are past their due date and need urgent attention.
     Only searches through open (non-closed) work packages.
-    
+
     Args:
         project_id: Optional project ID to filter by
         assignee_id: Optional user ID to filter by assignee
         priority_ids: Optional comma-separated priority IDs (e.g., "3" for high, or "3,4" for high+urgent)
         type_ids: Optional comma-separated type IDs (e.g., "1" for bugs, or "1,2" for bugs+features)
         page_size: Number of results to return (default: 50, max: 100)
-    
+
     Returns:
         Formatted list of overdue work packages sorted by most overdue first
-        
+
     Example:
         Find all high-priority overdue tasks assigned to user #5:
         {
@@ -935,9 +1068,9 @@ async def list_overdue_work_packages(
     """
     try:
         from datetime import date, datetime
-        
+
         client = get_client()
-        
+
         # Build filters list
         filters_list = [
             # Status must be open (not closed)
@@ -945,43 +1078,50 @@ async def list_overdue_work_packages(
             # Due date < today (overdue)
             # Note: OpenProject API doesn't support "<d" operator with single value
             # Workaround: Use "<>d" (between) with old start date and today
-            {"dueDate": {"operator": "<>d", "values": ["2000-01-01", date.today().isoformat()]}}
+            {
+                "dueDate": {
+                    "operator": "<>d",
+                    "values": ["2000-01-01", date.today().isoformat()],
+                }
+            },
         ]
-        
+
         # Add optional filters
         if assignee_id:
-            filters_list.append({"assignee": {"operator": "=", "values": [str(assignee_id)]}})
-        
+            filters_list.append(
+                {"assignee": {"operator": "=", "values": [str(assignee_id)]}}
+            )
+
         if priority_ids:
             # Parse comma-separated IDs
             priority_list = [p.strip() for p in priority_ids.split(",") if p.strip()]
             if priority_list:
-                filters_list.append({"priority": {"operator": "=", "values": priority_list}})
-        
+                filters_list.append(
+                    {"priority": {"operator": "=", "values": priority_list}}
+                )
+
         if type_ids:
             # Parse comma-separated IDs
             type_list = [t.strip() for t in type_ids.split(",") if t.strip()]
             if type_list:
                 filters_list.append({"type": {"operator": "=", "values": type_list}})
-        
+
         filters = json.dumps(filters_list)
-        
+
         # Validate page_size
         if page_size < 1 or page_size > 100:
             return format_error("page_size must be between 1 and 100")
-        
+
         result = await client.get_work_packages(
-            project_id=project_id,
-            filters=filters,
-            page_size=page_size
+            project_id=project_id, filters=filters, page_size=page_size
         )
-        
+
         work_packages = result.get("_embedded", {}).get("elements", [])
         total = result.get("total", 0)
-        
+
         if not work_packages:
             return "✅ No overdue work packages found!"
-        
+
         # Calculate days overdue for each task
         today = date.today()
         for wp in work_packages:
@@ -995,16 +1135,16 @@ async def list_overdue_work_packages(
                     wp["_days_overdue"] = 0
             else:
                 wp["_days_overdue"] = 0
-        
+
         # Sort by most overdue first
         work_packages.sort(key=lambda w: w.get("_days_overdue", 0), reverse=True)
-        
+
         # Format response
         text = f"⚠️ **Overdue Work Packages**: {total} task(s) past due date\n\n"
         text += format_work_package_list(work_packages, show_days_overdue=True)
-        
+
         return text
-        
+
     except Exception as e:
         return format_error(f"Failed to list overdue work packages: {str(e)}")
 
@@ -1015,23 +1155,23 @@ async def list_work_packages_due_soon(
     project_id: Optional[int] = None,
     assignee_id: Optional[int] = None,
     priority_ids: Optional[str] = None,
-    page_size: int = 50
+    page_size: int = 50,
 ) -> str:
     """List work packages due within the next N days.
-    
+
     This helps identify upcoming deadlines and prioritize work accordingly.
     Only searches through open (non-closed) work packages.
-    
+
     Args:
         days: Number of days to look ahead (default: 7)
         project_id: Optional project ID to filter by
         assignee_id: Optional user ID to filter by assignee
         priority_ids: Optional comma-separated priority IDs (e.g., "3,4")
         page_size: Number of results to return (default: 50)
-    
+
     Returns:
         Formatted list of work packages due soon, sorted by soonest first
-        
+
     Example:
         Show my tasks due in the next 3 days:
         {
@@ -1041,54 +1181,61 @@ async def list_work_packages_due_soon(
     """
     try:
         from datetime import date, timedelta, datetime
-        
+
         client = get_client()
-        
+
         # Validate days parameter
         if days < 1:
             return format_error("days must be at least 1")
         if days > 365:
             return format_error("days cannot exceed 365")
-        
+
         # Calculate date range
         today = date.today()
         target_date = today + timedelta(days=days)
-        
+
         # Build filters
         filters_list = [
             # Status must be open
             {"status": {"operator": "o", "values": []}},
             # Due date between today and target_date
-            {"dueDate": {"operator": "<>d", "values": [today.isoformat(), target_date.isoformat()]}}
+            {
+                "dueDate": {
+                    "operator": "<>d",
+                    "values": [today.isoformat(), target_date.isoformat()],
+                }
+            },
         ]
-        
+
         # Add optional filters
         if assignee_id:
-            filters_list.append({"assignee": {"operator": "=", "values": [str(assignee_id)]}})
-        
+            filters_list.append(
+                {"assignee": {"operator": "=", "values": [str(assignee_id)]}}
+            )
+
         if priority_ids:
             priority_list = [p.strip() for p in priority_ids.split(",") if p.strip()]
             if priority_list:
-                filters_list.append({"priority": {"operator": "=", "values": priority_list}})
-        
+                filters_list.append(
+                    {"priority": {"operator": "=", "values": priority_list}}
+                )
+
         filters = json.dumps(filters_list)
-        
+
         # Validate page_size
         if page_size < 1 or page_size > 100:
             return format_error("page_size must be between 1 and 100")
-        
+
         result = await client.get_work_packages(
-            project_id=project_id,
-            filters=filters,
-            page_size=page_size
+            project_id=project_id, filters=filters, page_size=page_size
         )
-        
+
         work_packages = result.get("_embedded", {}).get("elements", [])
         total = result.get("total", 0)
-        
+
         if not work_packages:
             return f"✅ No work packages due in the next {days} day(s)!"
-        
+
         # Calculate days until due
         for wp in work_packages:
             due_date_str = wp.get("dueDate")
@@ -1101,16 +1248,16 @@ async def list_work_packages_due_soon(
                     wp["_days_until"] = 999
             else:
                 wp["_days_until"] = 999
-        
+
         # Sort by soonest first
         work_packages.sort(key=lambda w: w.get("_days_until", 999))
-        
+
         # Format response
         text = f"⏰ **Work Packages Due Soon**: {total} task(s) due in next {days} day(s)\n\n"
         text += format_work_package_list(work_packages, show_days_until=True)
-        
+
         return text
-        
+
     except Exception as e:
         return format_error(f"Failed to list work packages due soon: {str(e)}")
 
@@ -1121,23 +1268,23 @@ async def list_unassigned_work_packages(
     priority_ids: Optional[str] = None,
     type_ids: Optional[str] = None,
     active_only: bool = True,
-    page_size: int = 50
+    page_size: int = 50,
 ) -> str:
     """List work packages that have no assignee.
-    
+
     This helps identify tasks that need to be assigned to team members.
     Useful for sprint planning and workload distribution.
-    
+
     Args:
         project_id: Optional project ID to filter by
         priority_ids: Optional comma-separated priority IDs (e.g., "3,4" for high+urgent)
         type_ids: Optional comma-separated type IDs (e.g., "1" for bugs only)
         active_only: If True, only show open work packages (default: True)
         page_size: Number of results to return (default: 50, max: 100)
-    
+
     Returns:
         Formatted list of unassigned work packages
-        
+
     Example:
         Find all unassigned high-priority bugs in project #5:
         {
@@ -1148,57 +1295,59 @@ async def list_unassigned_work_packages(
     """
     try:
         client = get_client()
-        
+
         # Build filters list
         filters_list = [
             # Assignee must be empty (unassigned)
             {"assignee": {"operator": "!*", "values": []}}
         ]
-        
+
         # Add status filter
         if active_only:
             filters_list.append({"status": {"operator": "o", "values": []}})
         else:
             filters_list.append({"status": {"operator": "*", "values": []}})
-        
+
         # Add optional filters
         if priority_ids:
             priority_list = [p.strip() for p in priority_ids.split(",") if p.strip()]
             if priority_list:
-                filters_list.append({"priority": {"operator": "=", "values": priority_list}})
-        
+                filters_list.append(
+                    {"priority": {"operator": "=", "values": priority_list}}
+                )
+
         if type_ids:
             type_list = [t.strip() for t in type_ids.split(",") if t.strip()]
             if type_list:
                 filters_list.append({"type": {"operator": "=", "values": type_list}})
-        
+
         filters = json.dumps(filters_list)
-        
+
         # Validate page_size
         if page_size < 1 or page_size > 100:
             return format_error("page_size must be between 1 and 100")
-        
+
         result = await client.get_work_packages(
-            project_id=project_id,
-            filters=filters,
-            page_size=page_size
+            project_id=project_id, filters=filters, page_size=page_size
         )
-        
+
         work_packages = result.get("_embedded", {}).get("elements", [])
         total = result.get("total", 0)
-        
+
         if not work_packages:
             return "✅ No unassigned work packages found!"
-        
+
         # Format response
         text = f"👤 **Unassigned Work Packages**: {total} task(s) without assignee\n\n"
         text += format_work_package_list(work_packages)
-        
+
         if total > page_size:
-            text += f"\n📄 Showing first {page_size} of {total} total unassigned tasks\n"
-        
+            text += (
+                f"\n📄 Showing first {page_size} of {total} total unassigned tasks\n"
+            )
+
         return text
-        
+
     except Exception as e:
         return format_error(f"Failed to list unassigned work packages: {str(e)}")
 
@@ -1210,12 +1359,12 @@ async def list_work_packages_created_recently(
     assignee_id: Optional[int] = None,
     type_ids: Optional[str] = None,
     active_only: bool = True,
-    page_size: int = 50
+    page_size: int = 50,
 ) -> str:
     """List work packages created in the last N days.
-    
+
     This helps identify new tasks and track task creation patterns.
-    
+
     Args:
         days: Number of days to look back (default: 7)
         project_id: Optional project ID to filter by
@@ -1223,10 +1372,10 @@ async def list_work_packages_created_recently(
         type_ids: Optional comma-separated type IDs (e.g., "1,2" for bugs+features)
         active_only: If True, only show open work packages (default: True)
         page_size: Number of results to return (default: 50, max: 100)
-    
+
     Returns:
         Formatted list of recently created work packages, sorted by newest first
-        
+
     Example:
         Show all bugs created in the last 3 days:
         {
@@ -1236,67 +1385,67 @@ async def list_work_packages_created_recently(
     """
     try:
         from datetime import date, timedelta, datetime
-        
+
         client = get_client()
-        
+
         # Validate days parameter
         if days < 1:
             return format_error("days must be at least 1")
         if days > 365:
             return format_error("days cannot exceed 365")
-        
+
         # Calculate date range
         # Note: Use <t operator for "ago" (created less than N days ago)
         filters_list = [
             # Created at < N days ago (i.e., within last N days)
             {"createdAt": {"operator": "<t", "values": [str(days)]}}
         ]
-        
+
         # Add status filter
         if active_only:
             filters_list.append({"status": {"operator": "o", "values": []}})
         else:
             filters_list.append({"status": {"operator": "*", "values": []}})
-        
+
         # Add optional filters
         if assignee_id:
-            filters_list.append({"assignee": {"operator": "=", "values": [str(assignee_id)]}})
-        
+            filters_list.append(
+                {"assignee": {"operator": "=", "values": [str(assignee_id)]}}
+            )
+
         if type_ids:
             type_list = [t.strip() for t in type_ids.split(",") if t.strip()]
             if type_list:
                 filters_list.append({"type": {"operator": "=", "values": type_list}})
-        
+
         filters = json.dumps(filters_list)
-        
+
         # Validate page_size
         if page_size < 1 or page_size > 100:
             return format_error("page_size must be between 1 and 100")
-        
+
         result = await client.get_work_packages(
-            project_id=project_id,
-            filters=filters,
-            page_size=page_size
+            project_id=project_id, filters=filters, page_size=page_size
         )
-        
+
         work_packages = result.get("_embedded", {}).get("elements", [])
         total = result.get("total", 0)
-        
+
         if not work_packages:
             return f"✅ No work packages created in the last {days} day(s)!"
-        
+
         # Sort by creation date (newest first)
         work_packages.sort(key=lambda w: w.get("createdAt", ""), reverse=True)
-        
+
         # Format response
         text = f"🆕 **Recently Created Work Packages**: {total} task(s) created in last {days} day(s)\n\n"
         text += format_work_package_list(work_packages)
-        
+
         if total > page_size:
             text += f"\n📄 Showing first {page_size} of {total} total\n"
-        
+
         return text
-        
+
     except Exception as e:
         return format_error(f"Failed to list recently created work packages: {str(e)}")
 
@@ -1307,31 +1456,31 @@ async def list_high_priority_work_packages(
     assignee_id: Optional[int] = None,
     type_ids: Optional[str] = None,
     active_only: bool = True,
-    page_size: int = 50
+    page_size: int = 50,
 ) -> str:
     """List work packages with high priority.
-    
+
     This tool finds tasks marked as high priority or urgent. Note that you need to know
     the priority ID for "High" in your OpenProject instance (typically 3 or 4).
     Use list_priorities tool first if you don't know the priority IDs.
-    
+
     Args:
         project_id: Optional project ID to filter by
         assignee_id: Optional user ID to filter by assignee
         type_ids: Optional comma-separated type IDs (e.g., "1" for bugs only)
         active_only: If True, only show open work packages (default: True)
         page_size: Number of results to return (default: 50, max: 100)
-    
+
     Returns:
         Formatted list of high priority work packages
-        
+
     Example:
         Show all high-priority bugs in project #5:
         {
             "project_id": 5,
             "type_ids": "1"
         }
-        
+
     Note:
         This assumes priority ID 3 = "High". If your instance uses different IDs,
         use list_priorities to find the correct ID, then use list_work_packages
@@ -1339,58 +1488,58 @@ async def list_high_priority_work_packages(
     """
     try:
         client = get_client()
-        
+
         # Build filters - assume priority ID 3 is "High"
         # Users can override by using list_work_packages with specific priority_ids
         filters_list = [
             # Priority = 3 (typically "High" in OpenProject)
             {"priority": {"operator": "=", "values": ["3"]}}
         ]
-        
+
         # Add status filter
         if active_only:
             filters_list.append({"status": {"operator": "o", "values": []}})
         else:
             filters_list.append({"status": {"operator": "*", "values": []}})
-        
+
         # Add optional filters
         if assignee_id:
-            filters_list.append({"assignee": {"operator": "=", "values": [str(assignee_id)]}})
-        
+            filters_list.append(
+                {"assignee": {"operator": "=", "values": [str(assignee_id)]}}
+            )
+
         if type_ids:
             type_list = [t.strip() for t in type_ids.split(",") if t.strip()]
             if type_list:
                 filters_list.append({"type": {"operator": "=", "values": type_list}})
-        
+
         filters = json.dumps(filters_list)
-        
+
         # Validate page_size
         if page_size < 1 or page_size > 100:
             return format_error("page_size must be between 1 and 100")
-        
+
         result = await client.get_work_packages(
-            project_id=project_id,
-            filters=filters,
-            page_size=page_size
+            project_id=project_id, filters=filters, page_size=page_size
         )
-        
+
         work_packages = result.get("_embedded", {}).get("elements", [])
         total = result.get("total", 0)
-        
+
         if not work_packages:
             return "✅ No high priority work packages found!"
-        
+
         # Format response
         text = f"🔴 **High Priority Work Packages**: {total} task(s)\n\n"
         text += "💡 Note: This lists tasks with priority ID 3 (typically 'High').\n"
         text += "   Use list_priorities to see all priority IDs in your instance.\n\n"
         text += format_work_package_list(work_packages)
-        
+
         if total > page_size:
             text += f"\n📄 Showing first {page_size} of {total} total\n"
-        
+
         return text
-        
+
     except Exception as e:
         return format_error(f"Failed to list high priority work packages: {str(e)}")
 
@@ -1401,23 +1550,23 @@ async def list_work_packages_nearly_complete(
     assignee_id: Optional[int] = None,
     min_percentage: int = 80,
     active_only: bool = True,
-    page_size: int = 50
+    page_size: int = 50,
 ) -> str:
     """List work packages that are nearly complete (high percentage done).
-    
+
     This helps identify tasks that are almost finished and may need a final push.
     Useful for sprint reviews and workload tracking.
-    
+
     Args:
         project_id: Optional project ID to filter by
         assignee_id: Optional user ID to filter by assignee
         min_percentage: Minimum completion percentage (default: 80, range: 1-99)
         active_only: If True, only show open work packages (default: True)
         page_size: Number of results to return (default: 50, max: 100)
-    
+
     Returns:
         Formatted list of nearly complete work packages
-        
+
     Example:
         Show tasks >90% complete in project #5:
         {
@@ -1427,52 +1576,52 @@ async def list_work_packages_nearly_complete(
     """
     try:
         client = get_client()
-        
+
         # Validate min_percentage
         if min_percentage < 1 or min_percentage > 99:
             return format_error("min_percentage must be between 1 and 99")
-        
+
         # Build filters
         filters_list = [
             # Percentage done >= min_percentage
             {"percentageDone": {"operator": ">=", "values": [str(min_percentage)]}}
         ]
-        
+
         # Add status filter
         if active_only:
             filters_list.append({"status": {"operator": "o", "values": []}})
         else:
             filters_list.append({"status": {"operator": "*", "values": []}})
-        
+
         # Add optional filters
         if assignee_id:
-            filters_list.append({"assignee": {"operator": "=", "values": [str(assignee_id)]}})
-        
+            filters_list.append(
+                {"assignee": {"operator": "=", "values": [str(assignee_id)]}}
+            )
+
         filters = json.dumps(filters_list)
-        
+
         # Validate page_size
         if page_size < 1 or page_size > 100:
             return format_error("page_size must be between 1 and 100")
-        
+
         result = await client.get_work_packages(
-            project_id=project_id,
-            filters=filters,
-            page_size=page_size
+            project_id=project_id, filters=filters, page_size=page_size
         )
-        
+
         work_packages = result.get("_embedded", {}).get("elements", [])
         total = result.get("total", 0)
-        
+
         if not work_packages:
             return f"✅ No work packages with ≥{min_percentage}% completion found!"
-        
+
         # Sort by percentage done (highest first)
         work_packages.sort(key=lambda w: w.get("percentageDone", 0), reverse=True)
-        
+
         # Format response
         text = f"📊 **Nearly Complete Work Packages**: {total} task(s) ≥{min_percentage}% done\n\n"
         text += format_work_package_list(work_packages)
-        
+
         # Add completion percentages in summary
         if work_packages:
             text += "\n**Completion Summary**:\n"
@@ -1482,13 +1631,11 @@ async def list_work_packages_nearly_complete(
                 text += f"  - #{wp.get('id')}: {percentage}% - {subject}\n"
             if len(work_packages) > 10:
                 text += f"  ... and {len(work_packages) - 10} more\n"
-        
+
         if total > page_size:
             text += f"\n📄 Showing first {page_size} of {total} total\n"
-        
+
         return text
-        
+
     except Exception as e:
         return format_error(f"Failed to list nearly complete work packages: {str(e)}")
-
-
diff --git a/src/utils/inputs.py b/src/utils/inputs.py
new file mode 100644
index 0000000..2152779
--- /dev/null
+++ b/src/utils/inputs.py
@@ -0,0 +1,56 @@
+"""Tolerant handling of wrapped Pydantic tool inputs.
+
+Several tools take a single parameter -- a Pydantic model named ``input``.
+FastMCP validates that argument against the model *before* the tool body runs.
+
+MCP clients differ in how they serialize structured arguments. Spec-compliant
+clients send ``input`` as a JSON object and these tools work unchanged. Other
+clients serialize object-typed arguments as a JSON *string* when the parameter's
+schema is not an explicit ``type: object``; FastMCP then receives a ``str`` and
+rejects it with a Pydantic ``model_type`` error before the body runs.
+
+``CoercibleModel`` makes any input model accept that JSON-string encoding in
+addition to a dict or model instance, so the server works with both kinds of
+client. ``as_model`` is the equivalent helper for non-decorated call paths.
+"""
+
+from __future__ import annotations
+
+import json
+from typing import Type, TypeVar, Union
+
+from pydantic import BaseModel, model_validator
+
+ModelT = TypeVar("ModelT", bound=BaseModel)
+
+
+class CoercibleModel(BaseModel):
+    """Input-model base that also accepts a JSON string.
+
+    A ``model_validator(mode="before")`` runs prior to field validation. When the
+    incoming value is a ``str`` (a client that serialized the object as JSON) it
+    is parsed into a dict first. Dicts and model instances pass through unchanged,
+    so spec-compliant clients are unaffected. Invalid JSON raises and surfaces as
+    a normal validation error.
+    """
+
+    @model_validator(mode="before")
+    @classmethod
+    def _accept_json_string(cls, data):
+        if isinstance(data, str):
+            return json.loads(data)
+        return data
+
+
+def as_model(model_cls: Type[ModelT], value: Union[ModelT, dict, str]) -> ModelT:
+    """Coerce a model instance, dict, or JSON string into ``model_cls``."""
+    if isinstance(value, model_cls):
+        return value
+    if isinstance(value, str):
+        return model_cls.model_validate_json(value)
+    if isinstance(value, dict):
+        return model_cls.model_validate(value)
+    raise TypeError(
+        f"{model_cls.__name__} input must be a model, dict, or JSON string, "
+        f"got {type(value).__name__}"
+    )
diff --git a/test_tools.py b/test_tools.py
index 9af9d49..562eab2 100644
--- a/test_tools.py
+++ b/test_tools.py
@@ -77,34 +77,22 @@ async def test_all_tools():
         print(f"FAIL FAILED: {e}")
 
     # Test 5: Create Work Package (CRITICAL) - DRY RUN
-    print("\n[5] Test: create_work_package validation (CRITICAL)")
+    print("\n[5] Test: create_work_package (CRITICAL)")
     try:
-        from src.tools.work_packages import CreateWorkPackageInput
-        # Validate input model only (don't actually create)
-        test_input = CreateWorkPackageInput(
-            project_id=1,
-            subject="Test Work Package",
-            type_id=1,
-            description="This is a test"
-        )
-        print(f"OK Input validation PASSED")
-        print(f"   Model: {test_input.model_dump()}")
+        # create_work_package now uses flat parameters (no wrapped input model)
+        from src.tools.work_packages import create_work_package
+        assert create_work_package is not None
+        print("OK create_work_package importable (flat params)")
     except Exception as e:
         print(f"FAIL FAILED: {e}")
 
     # Test 6: Update Work Package (CRITICAL) - DRY RUN
-    print("\n[6] Test: update_work_package validation (CRITICAL)")
+    print("\n[6] Test: update_work_package (CRITICAL)")
     try:
-        from src.tools.work_packages import UpdateWorkPackageInput
-        # Validate input model only (don't actually update)
-        test_input = UpdateWorkPackageInput(
-            work_package_id=123,
-            status_id=5,
-            assignee_id=7,
-            percentage_done=50
-        )
-        print(f"OK Input validation PASSED")
-        print(f"   Model: {test_input.model_dump()}")
+        # update_work_package now uses flat parameters (no wrapped input model)
+        from src.tools.work_packages import update_work_package
+        assert update_work_package is not None
+        print("OK update_work_package importable (flat params)")
     except Exception as e:
         print(f"FAIL FAILED: {e}")
 
diff --git a/tests/conftest.py b/tests/conftest.py
new file mode 100644
index 0000000..2a35fa3
--- /dev/null
+++ b/tests/conftest.py
@@ -0,0 +1,11 @@
+"""Shared test setup.
+
+Importing the tool modules pulls in ``src.server``, which requires OpenProject
+credentials to construct its client. The tests never hit the network, so we set
+dummy credentials before anything imports the server.
+"""
+
+import os
+
+os.environ.setdefault("OPENPROJECT_URL", "https://openproject.test")
+os.environ.setdefault("OPENPROJECT_API_KEY", "dummy-key-for-tests")
diff --git a/tests/test_inputs.py b/tests/test_inputs.py
new file mode 100644
index 0000000..0575cbc
--- /dev/null
+++ b/tests/test_inputs.py
@@ -0,0 +1,117 @@
+"""Tests for tolerant input handling (CoercibleModel + as_model).
+
+These verify the server accepts both the object and the JSON-string encoding of
+wrapped tool inputs, so it works regardless of how an MCP client serializes
+structured arguments.
+"""
+
+import importlib
+import inspect
+import pkgutil
+from typing import Optional
+
+import pytest
+from pydantic import BaseModel
+
+from src.utils.inputs import CoercibleModel, as_model
+
+
+class Demo(CoercibleModel):
+    project_id: int
+    subject: str
+    type_id: int
+    description: Optional[str] = None
+
+
+# --- CoercibleModel: accepts object and JSON-string encodings ---
+
+
+def test_coercible_accepts_dict():
+    m = Demo.model_validate({"project_id": 1, "subject": "x", "type_id": 3})
+    assert m.project_id == 1 and m.type_id == 3
+
+
+def test_coercible_accepts_json_string():
+    # This is the encoding that previously failed with a `model_type` error.
+    m = Demo.model_validate('{"project_id": 1, "subject": "x", "type_id": 3}')
+    assert m.project_id == 1 and m.subject == "x"
+
+
+def test_coercible_as_nested_field_accepts_json_string():
+    # Mirrors how FastMCP validates a wrapped `input` parameter.
+    class Call(BaseModel):
+        input: Demo
+
+    c = Call.model_validate(
+        {"input": '{"project_id": 2, "subject": "y", "type_id": 4}'}
+    )
+    assert isinstance(c.input, Demo) and c.input.project_id == 2
+
+
+def test_coercible_invalid_json_raises():
+    with pytest.raises(Exception):
+        Demo.model_validate("{not valid json")
+
+
+# --- as_model helper ---
+
+
+@pytest.mark.parametrize(
+    "value",
+    [
+        {"project_id": 1, "subject": "x", "type_id": 3},
+        '{"project_id": 1, "subject": "x", "type_id": 3}',
+        Demo(project_id=1, subject="x", type_id=3),
+    ],
+)
+def test_as_model_accepts_all_encodings(value):
+    m = as_model(Demo, value)
+    assert m.project_id == 1 and m.type_id == 3
+
+
+def test_as_model_rejects_bad_type():
+    with pytest.raises(TypeError):
+        as_model(Demo, 123)
+
+
+def test_as_model_rejects_invalid_payload():
+    with pytest.raises(Exception):
+        as_model(Demo, '{"subject": "missing required fields"}')
+
+
+# --- Tripwire: every wrapped *Input model must be coercible ---
+
+
+def _iter_input_models():
+    import src.tools as tools_pkg
+
+    for mod in pkgutil.iter_modules(tools_pkg.__path__):
+        module = importlib.import_module(f"src.tools.{mod.name}")
+        for name, obj in inspect.getmembers(module, inspect.isclass):
+            if (
+                name.endswith("Input")
+                and issubclass(obj, BaseModel)
+                and obj.__module__ == module.__name__
+            ):
+                yield f"{mod.name}.{name}", obj
+
+
+def test_every_wrapped_input_model_is_coercible():
+    offenders = [
+        qualname
+        for qualname, model in _iter_input_models()
+        if not issubclass(model, CoercibleModel)
+    ]
+    assert not offenders, (
+        "These *Input models must subclass CoercibleModel so the server tolerates "
+        f"JSON-string arguments: {offenders}"
+    )
+
+
+def test_flattened_work_package_tools_have_no_wrapped_input():
+    # The two hot tools were flattened to explicit params; guard against a
+    # regression back to a single wrapped `input` model.
+    import src.tools.work_packages as wp
+
+    assert not hasattr(wp, "CreateWorkPackageInput")
+    assert not hasattr(wp, "UpdateWorkPackageInput")
diff --git a/tests/test_integration_smoke.py b/tests/test_integration_smoke.py
new file mode 100644
index 0000000..5e378e0
--- /dev/null
+++ b/tests/test_integration_smoke.py
@@ -0,0 +1,84 @@
+"""Live integration smoke test for the flattened work-package tools.
+
+Skipped by default. It exercises the real flattened payload-building end to end
+(create -> read-back -> update(status) -> delete) against a live OpenProject
+instance, so the change is covered beyond signatures/imports. It cleans up the
+work package it creates.
+
+There is intentionally no standalone ``get_work_package`` tool in this server, so
+the read-back uses ``search_work_packages`` (which matches by subject or ID). The
+tool imports are at module top so a missing/renamed tool fails collection even
+when this test is skipped.
+
+Enable by pointing at a disposable test project with real credentials:
+
+    OPENPROJECT_SMOKE=1 \\
+    OPENPROJECT_URL=https://your.instance \\
+    OPENPROJECT_API_KEY=... \\
+    OPENPROJECT_TEST_PROJECT_ID=458 \\
+    OPENPROJECT_TEST_TYPE_ID=1 \\
+    OPENPROJECT_TEST_STATUS_ID=... \\  # optional; exercises the status update
+    uv run pytest tests/test_integration_smoke.py -v
+"""
+
+import os
+import re
+
+import pytest
+
+from src.tools.work_packages import (
+    create_work_package,
+    update_work_package,
+    search_work_packages,
+    delete_work_package,
+)
+
+SMOKE = os.getenv("OPENPROJECT_SMOKE") == "1"
+PROJECT_ID = os.getenv("OPENPROJECT_TEST_PROJECT_ID")
+TYPE_ID = os.getenv("OPENPROJECT_TEST_TYPE_ID")
+STATUS_ID = os.getenv("OPENPROJECT_TEST_STATUS_ID")
+
+pytestmark = pytest.mark.skipif(
+    not (SMOKE and PROJECT_ID and TYPE_ID),
+    reason=(
+        "live smoke test disabled; set OPENPROJECT_SMOKE=1 + "
+        "OPENPROJECT_TEST_PROJECT_ID + OPENPROJECT_TEST_TYPE_ID (and real creds)"
+    ),
+)
+
+
+def _wp_id(text: str) -> int:
+    match = re.search(r"#(\d+)", text)
+    assert match, f"could not parse work package id from: {text!r}"
+    return int(match.group(1))
+
+
+async def test_work_package_crud_roundtrip():
+    # @mcp.tool wraps the coroutine in a FunctionTool; the callable is at `.fn`.
+    create = create_work_package.fn
+    update = update_work_package.fn
+    search = search_work_packages.fn
+    delete = delete_work_package.fn
+
+    created = await create(
+        project_id=int(PROJECT_ID),
+        subject="[smoke] flattened create/update payload",
+        type_id=int(TYPE_ID),
+        description="Temporary work package from the integration smoke test.",
+    )
+    assert "created successfully" in created.lower(), created
+    wp_id = _wp_id(created)
+
+    try:
+        # No standalone get_work_package tool exists; read back via search by id.
+        found = await search(query=str(wp_id), project_id=int(PROJECT_ID))
+        assert str(wp_id) in found, found
+
+        kwargs = {"work_package_id": wp_id, "subject": "[smoke] updated subject"}
+        if STATUS_ID:
+            kwargs["status_id"] = int(STATUS_ID)
+        updated = await update(**kwargs)
+        assert "updated successfully" in updated.lower(), updated
+    finally:
+        deleted = await delete(wp_id)
+        assert "deleted successfully" in deleted.lower(), deleted
diff --git a/tests/test_update_readonly.py b/tests/test_update_readonly.py
new file mode 100644
index 0000000..592f560
--- /dev/null
+++ b/tests/test_update_readonly.py
@@ -0,0 +1,95 @@
+"""Unit coverage for update_work_package's read-only percentage_done handling.
+
+On some OpenProject instances percentage_done is read-only and the API rejects it
+with HTTP 422 PropertyIsReadOnly. The tool should drop that field and retry so the
+rest of the update still applies, without swallowing unrelated errors. These tests
+mock the client, so they are network-free.
+"""
+
+from src.tools import work_packages as wp
+
+
+class _FakeClient:
+    def __init__(self):
+        self.calls = []
+
+    async def update_work_package(self, wp_id, data):
+        self.calls.append(dict(data))
+        if "percentage_done" in data:
+            raise Exception(
+                "API Error 422: PropertyIsReadOnly - "
+                "Percentage done was attempted to be written but is not writable."
+            )
+        return {"id": wp_id, "subject": "S", "_embedded": {}}
+
+
+async def test_percentage_readonly_retries_without_field(monkeypatch):
+    fake = _FakeClient()
+    monkeypatch.setattr(wp, "get_client", lambda: fake)
+
+    result = await wp.update_work_package.fn(
+        work_package_id=1, status_id=2, percentage_done=50
+    )
+
+    assert "updated successfully" in result.lower()
+    assert "percentage_done is read-only" in result
+    # First attempt included the field; retry dropped it but kept status_id.
+    assert len(fake.calls) == 2
+    assert "percentage_done" in fake.calls[0]
+    assert "percentage_done" not in fake.calls[1]
+    assert fake.calls[1]["status_id"] == 2
+
+
+async def test_percentage_readonly_only_field_returns_clear_message(monkeypatch):
+    fake = _FakeClient()
+    monkeypatch.setattr(wp, "get_client", lambda: fake)
+
+    result = await wp.update_work_package.fn(work_package_id=1, percentage_done=50)
+
+    assert "read-only" in result.lower()
+    assert "no other fields" in result.lower()
+    assert len(fake.calls) == 1  # no pointless retry with an empty payload
+
+
+async def test_non_readonly_error_is_not_swallowed(monkeypatch):
+    class _BoomClient:
+        async def update_work_package(self, wp_id, data):
+            raise Exception("API Error 404: not found")
+
+    monkeypatch.setattr(wp, "get_client", lambda: _BoomClient())
+
+    result = await wp.update_work_package.fn(
+        work_package_id=1, status_id=2, percentage_done=50
+    )
+
+    assert "failed to update" in result.lower()
+
+
+async def test_percentage_readonly_detected_via_identifier_when_localized(monkeypatch):
+    # Even with a translated/reworded human message, detection must still fire off
+    # the stable v3 errorIdentifier + attribute name in the 422 body.
+    class _LocalizedClient:
+        def __init__(self):
+            self.calls = []
+
+        async def update_work_package(self, wp_id, data):
+            self.calls.append(dict(data))
+            if "percentage_done" in data:
+                raise Exception(
+                    'API Error 422: {"_type":"Error","errorIdentifier":'
+                    '"urn:openproject-org:api:v3:errors:PropertyIsReadOnly",'
+                    '"message":"Fertigstellungsgrad ist schreibgeschuetzt.",'
+                    '"_embedded":{"details":{"attribute":"percentageDone"}}}'
+                )
+            return {"id": wp_id, "subject": "S", "_embedded": {}}
+
+    fake = _LocalizedClient()
+    monkeypatch.setattr(wp, "get_client", lambda: fake)
+
+    result = await wp.update_work_package.fn(
+        work_package_id=1, status_id=2, percentage_done=50
+    )
+
+    assert "updated successfully" in result.lower()
+    assert len(fake.calls) == 2
+    assert "percentage_done" not in fake.calls[1]
diff --git a/uv.lock b/uv.lock
index 6a79137..8777d00 100644
--- a/uv.lock
+++ b/uv.lock
@@ -164,6 +164,15 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/f8/aa/5082412d1ee302e9e7d80b6949bc4d2a8fa1149aaab610c5fc24709605d6/authlib-1.6.5-py2.py3-none-any.whl", hash = "sha256:3e0e0507807f842b02175507bdee8957a1d5707fd4afb17c32fb43fee90b6e3a", size = 243608, upload-time = "2025-10-02T13:36:07.637Z" },
 ]
 
+[[package]]
+name = "backports-asyncio-runner"
+version = "1.2.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/8e/ff/70dca7d7cb1cbc0edb2c6cc0c38b65cba36cccc491eca64cabd5fe7f8670/backports_asyncio_runner-1.2.0.tar.gz", hash = "sha256:a5aa7b2b7d8f8bfcaa2b57313f70792df84e32a2a746f585213373f900b42162", size = 69893, upload-time = "2025-07-02T02:27:15.685Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/a0/59/76ab57e3fe74484f48a53f8e337171b4a2349e506eabe136d7e01d059086/backports_asyncio_runner-1.2.0-py3-none-any.whl", hash = "sha256:0da0a936a8aeb554eccb426dc55af3ba63bcdc69fa1a600b5bb305413a4477b5", size = 12313, upload-time = "2025-07-02T02:27:14.263Z" },
+]
+
 [[package]]
 name = "beartype"
 version = "0.22.7"
@@ -998,6 +1007,7 @@ dev = [
     { name = "black" },
     { name = "flake8" },
     { name = "pytest" },
+    { name = "pytest-asyncio" },
 ]
 
 [package.metadata]
@@ -1005,11 +1015,12 @@ requires-dist = [
     { name = "aiohttp", specifier = ">=3.8.0" },
     { name = "black", marker = "extra == 'dev'", specifier = ">=22.0.0" },
     { name = "certifi", specifier = ">=2022.0.0" },
-    { name = "fastmcp", specifier = ">=2.0.0" },
+    { name = "fastmcp", specifier = ">=2.0.0,<3" },
     { name = "flake8", marker = "extra == 'dev'", specifier = ">=4.0.0" },
     { name = "mcp", specifier = ">=1.0.0" },
     { name = "pydantic", specifier = ">=2.0.0" },
     { name = "pytest", marker = "extra == 'dev'", specifier = ">=7.0.0" },
+    { name = "pytest-asyncio", marker = "extra == 'dev'", specifier = ">=0.21.0" },
     { name = "python-dotenv", specifier = ">=1.0.0" },
     { name = "starlette", specifier = ">=0.27.0" },
     { name = "uvicorn", specifier = ">=0.24.0" },
@@ -1392,6 +1403,20 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/a8/a4/20da314d277121d6534b3a980b29035dcd51e6744bd79075a6ce8fa4eb8d/pytest-8.4.2-py3-none-any.whl", hash = "sha256:872f880de3fc3a5bdc88a11b39c9710c3497a547cfa9320bc3c5e62fbf272e79", size = 365750, upload-time = "2025-09-04T14:34:20.226Z" },
 ]
 
+[[package]]
+name = "pytest-asyncio"
+version = "1.4.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "backports-asyncio-runner", marker = "python_full_version < '3.11'" },
+    { name = "pytest" },
+    { name = "typing-extensions", marker = "python_full_version < '3.13'" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/43/7c/d36d04db312ecf4298932ef77e6e4a9e8ad017906e24e34f0b0c361a2473/pytest_asyncio-1.4.0.tar.gz", hash = "sha256:c6c0d2259945122819f171a32ecea2c349ead889ee28176caaf492143424be42", size = 58514, upload-time = "2026-05-26T09:56:04.083Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/03/e2/08a497ef684b88559c9cc5f4ad53a37e7b99e727094a86d6ea32536d5d3c/pytest_asyncio-1.4.0-py3-none-any.whl", hash = "sha256:933ca923a23075a87fb7070c0ec272a6848489824d887c85c812670932835aa1", size = 16930, upload-time = "2026-05-26T09:56:02.576Z" },
+]
+
 [[package]]
 name = "python-dotenv"
 version = "1.1.1"