a digital entity named phi that roams bsky
phi.zzstoatzz.io
pre-enter MCP servers before agent.run() to fix cancel scope error
fresh instances per run alone isn't enough — pydantic-ai runs parallel
tool calls via asyncio.create_task, and each does async with self: on
the MCP server. the first task opens the connection (anyio task group),
the last to finish closes it in a different task → RuntimeError.
fix: enter MCP servers in the calling task via AsyncExitStack before
agent.run(). parallel tool calls then only bump the reference count
(1→2→...→1) without ever hitting 0, so the connection open/close
both happen in the same task.
ref: https://github.com/pydantic/pydantic-ai/issues/2818
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
+10
-1
+10
-1
src/bot/agent.py
+10
-1
src/bot/agent.py
···
1
1
"""MCP-enabled agent for phi with structured memory."""
2
2
3
3
import asyncio
4
+
import contextlib
4
5
import ipaddress
5
6
import logging
6
7
import os
···
408
409
memory=self.memory,
409
410
thread_uri=thread_uri,
410
411
)
411
-
result = await self.agent.run(user_prompt, deps=deps, toolsets=self._mcp_toolsets())
412
+
# Enter MCP servers before agent.run() so the connection is opened
413
+
# in this task. Parallel tool calls inside agent.run() then just bump
414
+
# the reference count instead of opening/closing across tasks.
415
+
# https://github.com/pydantic/pydantic-ai/issues/2818
416
+
toolsets = self._mcp_toolsets()
417
+
async with contextlib.AsyncExitStack() as stack:
418
+
for ts in toolsets:
419
+
await stack.enter_async_context(ts)
420
+
result = await self.agent.run(user_prompt, deps=deps, toolsets=toolsets)
412
421
logger.info(f"agent decided: {result.output.action}" + (f" - {result.output.text[:80]}" if result.output.text else "") + (f" ({result.output.reason})" if result.output.reason else ""))
413
422
414
423
# Store interaction and extract observations