Motivation and Context

StreamableHTTPTransport is not handling connection/request errors properly, which leads to cancellation of the whole anyio.TaskGroup.

The fix should make connection/setup failures behave like normal request errors instead of letting them tear down the transport task group in a way that triggers cancellation churn.

Without the fix, a request collapses into an infinite AnyIO cancellation retry loop which may cause a sustained high CPU use.

How Has This Been Tested?

I was primarily testing this with Agno and FastMCP. There are some changes I had to make to Agno itself as well.

Here's how I've tested it:

1. Start FastMCP server
2. Start HTTP app running Agno
3. Chat with the agent; ask it to use tools; everything good so far
4. Kill MCP server
5. Ask the agent to run tool
6. MCP client fails and I end up with a cancellation retry loop

Breaking Changes

No code changes or configuration is needed.

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Documentation update

Checklist

• I have read the MCP Documentation
• My code follows the repository's style guidelines
• New and existing tests pass locally
• I have added appropriate error handling
• I have added or updated documentation as needed

Additional context