fix(mcp): send search via POST and preflight oversized store payloads

tests/test_mcp.py

@@ -107,6 +107,13 @@ def test_store_error(mcp_server):
     assert "Embedding failed" in result
 
 
+def test_store_rejects_oversized_content_without_http(mcp_server):
+    with patch.object(mcp_server.urllib.request, "urlopen") as mocked:
+        result = mcp_server.memory_store("x" * (mcp_server._MAX_COMMIT_TEXT_CHARS + 1))
+    assert "content too long" in result
+    mocked.assert_not_called()
+
+
 # -- memory_search ----------------------------------------------------------
 
 
@@ -149,20 +156,24 @@ def test_search_empty(mcp_server):
     assert result == "No matching memories found."
 
 
-def test_search_limit_clamped(mcp_server):
-    captured_url = {}
+def test_search_limit_clamped_and_uses_post_body(mcp_server):
+    captured = {}
 
     def fake_urlopen(req, **kwargs):
-        captured_url["url"] = req.full_url
+        captured["url"] = req.full_url
+        captured["method"] = req.get_method()
+        captured["data"] = json.loads(req.data.decode())
         return _make_response({"results": [], "elapsed_ms": 0})
 
     with patch.object(mcp_server.urllib.request, "urlopen", side_effect=fake_urlopen):
         mcp_server.memory_search("q", limit=999)
-    assert "limit=30" in captured_url["url"]
+    assert captured["url"] == "http://test:7777/search"
+    assert captured["method"] == "POST"
+    assert captured["data"]["limit"] == 30
 
     with patch.object(mcp_server.urllib.request, "urlopen", side_effect=fake_urlopen):
         mcp_server.memory_search("q", limit=-5)
-    assert "limit=1" in captured_url["url"]
+    assert captured["data"]["limit"] == 1
 
 
 def test_search_score_fallback_chain(mcp_server):
@@ -375,12 +386,16 @@ def test_collection_params_custom_bank(mcp_server, monkeypatch):
 
 def test_search_includes_collection_param(mcp_server, monkeypatch):
     monkeypatch.setattr(mcp_server, "BANK_ID", "custom-bank")
-    captured_url = {}
+    captured = {}
 
     def fake_urlopen(req, **kwargs):
-        captured_url["url"] = req.full_url
+        captured["url"] = req.full_url
+        captured["method"] = req.get_method()
+        captured["data"] = json.loads(req.data.decode())
         return _make_response({"results": [], "elapsed_ms": 0})
 
     with patch.object(mcp_server.urllib.request, "urlopen", side_effect=fake_urlopen):
         mcp_server.memory_search("test")
-    assert "collection=custom-bank" in captured_url["url"]
+    assert captured["url"] == "http://test:7777/search"
+    assert captured["method"] == "POST"
+    assert captured["data"]["collection"] == "custom-bank"

tools/mcp/server.py

@@ -15,14 +15,15 @@
 import json
 import logging
 import os
-import urllib.parse
 import urllib.request
 from typing import Any
 
 from fastmcp import FastMCP
 
 logger = logging.getLogger("rasputin.mcp")
 
+_MAX_COMMIT_TEXT_CHARS = 8000
+
 RASPUTIN_URL = os.environ.get("RASPUTIN_URL", "http://127.0.0.1:7777")
 RASPUTIN_TOKEN = os.environ.get("RASPUTIN_TOKEN", "")
 BANK_ID = os.environ.get("RASPUTIN_BANK_ID", "")
@@ -91,6 +92,9 @@ def memory_store(
         importance: Priority 0-100.  Default 60.  Use 80+ for critical
                     decisions, 40 for background context.
     """
+    if len(content) > _MAX_COMMIT_TEXT_CHARS:
+        return f"Error: content too long (max {_MAX_COMMIT_TEXT_CHARS} characters)."
+
     payload: dict[str, Any] = {
         "text": content,
         "source": source,
@@ -133,10 +137,9 @@ def memory_search(
         limit: Maximum results to return (1-30, default 10).
     """
     limit = max(1, min(30, limit))
-    qs: dict[str, Any] = {"q": query, "limit": limit}
-    qs.update(_collection_params())
-    params = urllib.parse.urlencode(qs)
-    result = _api(f"/search?{params}")
+    payload: dict[str, Any] = {"q": query, "limit": limit}
+    payload.update(_collection_params())
+    result = _api("/search", method="POST", data=payload)
     results = result.get("results", [])
     if not results:
         return "No matching memories found."