[fix] mcp chat completions

[uc4w6c]

Relevant issues

#N/A

https://www.loom.com/share/c05d94351ed94dcd8f03d3d64bce5393

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• I have Added testing in the tests/litellm/ directory, Adding at least 1 test is a hard requirement - see details
• My PR passes all unit tests on make test-unit
• My PR's scope is as isolated as possible, it only solves 1 specific problem

CI (LiteLLM team)

• Branch creation CI run
  Link:

• CI run for the last commit
  Link:

• Merge / cherry-pick CI run
  Links:

Type

🐛 Bug Fix
✅ Test

Changes

Fix MCP Chat Completions Streaming Error

Problem

When using MCP tools with /chat/completions endpoint with stream: true, the request failed with:

RuntimeError: Timeout context manager should be used inside a task
litellm.APIConnectionError: Timeout context manager should be used inside a task

Example request that failed:

bash curl -X POST http://localhost:4000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-1234" \
  -d '{
    "model": "claude-sonnet-4",
    "messages": [
      {
        "role": "user",
        "content": "use echo tool"}
    ],
    "tools": [
      {
        "type": "mcp",
        "server_url": "litellm_proxy/mcp/everything",
        "server_label": "everything",
        "require_approval": "never"
      }
    ],
    "stream": true
}'

Root Cause

The MCP handler implementation created nested event loops:
acompletion (async context)
→ run_in_executor → completion (new thread)
→ run_async_function (NEW EVENT LOOP in new thread)
→ handle_chat_completion_with_mcp
→ acompletion (ANOTHER EVENT LOOP)
→ streaming response
❌ aiohttp timeout error

This multi-layer nesting caused aiohttp's timeout context manager to fail because it couldn't find a proper asyncio task context during streaming.

The issue only appeared during streaming (not in debug mode with slower execution) due to timing-sensitive asyncio task context resolution.

Solution

Aligned MCP chat completions with the responses API pattern to eliminate event loop nesting:

After fix:
acompletion (async context)
→ run_in_executor → completion (new thread)
→ MCP check → acompletion_with_mcp() coroutine
→ await coroutine in SAME event loop
→ streaming response
✅ works correctly

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Review	Updated (UTC)
litellm	Ready	Preview, Comment	Jan 15, 2026 8:16am