feat: publishing manifests for MCP registry, Claude Code, and Gemini CLI#284
Merged
Conversation
Phase 1 - CI/CD fixes and bug squashing:
- Fix .codecov.yml paths (agent-toolkit -> clio-kit) and add missing flags
- Remove continue-on-error from quality_control.yml for ruff/test/security
- Delete redundant test-mcps.yml workflow
- Fix docs-and-website.yml branch target (dev -> main)
- Fix pandas: add missing hypothesis_testing import
- Fix jarvis: remove dead mcp.run(transport="sse") call
- Fix compression: add decompress_file tool
Phase 2 - Structural standardization:
- Standardize all FastMCP() names to kebab-case matching directory
- Convert 12 flat-layout servers to proper src/{name}_mcp/ packages
- Standardize build backend to hatchling across all servers
- Standardize transport to stdio/http with argparse + env var fallback
Phase 3 - FastMCP 3.0 upgrade (all 16 servers):
- Bump fastmcp dependency to >=3.0.0rc2
- Add instructions= to all FastMCP() constructors
- Add annotations= (readOnlyHint, destructiveHint, idempotentHint) to all tools
- Add tags= to all tools for categorization
- Replace ad-hoc error dicts with raise ToolError() from fastmcp.exceptions
- Add @mcp.resource() (at least 1 per server)
- Add @mcp.prompt() with Message from fastmcp.prompts (at least 1 per server)
- Trim verbose tool descriptions to 1-2 sentences
- Use Annotated[type, Field(description=...)] for parameter docs
- Add list_page_size=10 for servers with 10+ tools
- Update all tests for v3 API (no more FunctionTool/.fn() patterns)
4A: Replace AST-based doc generation with FastMCP 3.0 metadata extraction - New scripts/extract_mcp_metadata.py imports servers via async API - generate_docs.py uses subprocess per-server instead of AST parsing - readme_filler.py rewritten to use real tool/resource/prompt metadata - Includes annotations, tags, and resource templates in extracted data 4B: Consolidate Python version test matrix - Merge 4 separate jobs (test, python-3-10, python-3-11, python-3-12) into single matrix job with python-version: ["3.10", "3.11", "3.12"] - Coverage upload only on 3.12 - continue-on-error only for 3.10 and 3.11 4C: Add FastMCP validation CI job - New validate-fastmcp job in quality_control.yml - scripts/validate_fastmcp.py checks: instructions, annotations, tags, resources (static + templates), and prompts per server 4D: Create reusable composite action for MCP setup - .github/actions/setup-mcp/action.yml handles uv + Python + deps - Used by ruff, mypy, test, security, and validate-fastmcp jobs
- parquet: add ToolError wrapper for handler error responses - paraview: add list_page_size=10 for catalog pagination (28 tools) - node-hardware: replace old isError/content/_meta dict with clean return - node-hardware: update test to match new health_check return format
- Update FastMCP version references (0.2.0 -> 3.0.0rc2) - Add FastMCP 3.0 server pattern with annotations, tags, ToolError, resources, prompts - Add key imports reference - Update "Adding a New MCP Server" with all required FastMCP 3.0 features - Update CI/CD section with validate-fastmcp job and composite action - Add new files to Important Files Reference - Add paraview to server list (was missing)
…data Updated tool counts, descriptions, and metadata from FastMCP 3.0 async API. All 16 servers now reflected with correct tool/resource/prompt counts.
…ini CLI Add server.json (MCP registry), .claude-plugin/plugin.json and .mcp.json (Claude Code marketplace), gemini-extension.json (Gemini CLI extension), and claude_desktop_config.json (Claude Desktop reference config) for all 16 MCP servers. - New generate_server_json.py script extracts live FastMCP metadata and produces all manifest files in one pass - Update publish.yml with MCP registry (mcp-publisher) and Smithery jobs - Update quality_control.yml with server.json validation job - Update readme_filler.py with correct Gemini CLI and Claude Code sections - Fix pandas/parquet .gitignore to allow publishing JSON files
- Remove unused imports (F401): asyncio, sys, os, ToolError across arxiv, chronolog, node-hardware, parallel-sort, plot, slurm - Apply ruff format to all server and test files - Add S104 to ndp ruff ignore (intentional 0.0.0.0 binding for HTTP)
Contributor
|
Awesome. I love it. Particularly the one-sweep to upgrade and standardize all servers. Thanks man. I also liked the publishing as packages cleanup. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
io.github.iowarp/{name}-mcp.claude-plugin/marketplace.jsonat repo root + per-server.claude-plugin/plugin.jsonand.mcp.json(install viaclaude mcp add clio-{name} -- uvx clio-kit {name})gemini-extension.jsonbundles all 16 servers (install viagemini extensions install)claude_desktop_config.jsonreference file at repo rootvalidate-server-jsonjob in quality_control.yml,publish-to-mcp-registryandpublish-to-smitheryjobs in publish.ymlscripts/generate_server_json.py— single script generates all manifests from live FastMCP metadatascripts/readme_filler.py— adds Claude Code section, fixes Gemini section (was using nonexistent--mcpflag).gitignoreupdated to allow publishing JSON filesFiles added/modified (57 total)
Test plan
ruff checkandruff format --checkpass on all modified scriptsvalidate-server-jsonCI logic tested locally on compression and hdf5Manual steps needed after merge
SMITHERY_TOKENGitHub secret for Smithery publishingpublish.ymlworkflow