If using MCP with docker and docker isn't running, process hangs

[njbrake]

I was running integration tests and the agent was hanging:

tests/integration/test_load_and_run_multi_agent.py ✓✓✓ 36% ███▋ [04/22/25 09:18:05] INFO Final answer: The best agent framework depends on your specific needs, but some of the top options include Microsoft AutoGen for automation and multi-agent collaboration, OpenAI Swarm for lightweight multi-agent user_interaction.py:22
orchestration, and various Bridgged frameworks. For tailored recommendations, consider your use case and requirements.
tests/integration/test_load_and_run_multi_agent.py ✓✓✓✓s✓ 48% ████▊ docker: Cannot connect to the Docker daemon at unix:///Users/nbrake/.docker/run/docker.sock. Is the docker daemon running?.
See 'docker run --help'.

We need to figure out and make the library raise an error and exit if there are failures in setting up the MCP connections

[njbrake]

We could hack around it on our side, but I think this needs to go into each framework.

Per https://modelcontextprotocol.io/specification/draft/basic/lifecycle#timeouts

"Implementations SHOULD establish timeouts for all sent requests, to prevent hung connections and resource exhaustion."

[njbrake]

First up, Llama Index run-llama/llama_index#18512

[njbrake]

Next, OpenAI openai/openai-agents-python#580

[njbrake]

Google ADk google/adk-python#364

[njbrake]

Made an issue for agno agno-agi/agno#2953

[njbrake]

MCPAdapt (which is the thing that Smolagents uses) grll/mcpadapt#35