What
In-process SDK MCP servers can no longer wait forever while connecting. These are add-on tool servers that a program built on the Claude Code SDK runs inside its own process. The connection attempt is now limited by MCP_TIMEOUT, which defaults to 30000 milliseconds (30 seconds).
If the time runs out:
- The connection is closed and the server is reported as failed.
- An error says
SDK MCP server "<name>" connection timed out after <n>ms. - The failure is logged under its own reason,
mcp_sdk_connect_timeout.
Why
An SDK MCP server that never finishes connecting now shows up as a clear failure with a named server. Before, it could leave things waiting with no explanation. If a server of yours is slow to start, raising MCP_TIMEOUT gives it more time.
Names in the bundleMCP_TIMEOUT
The entry above is what we published on the day. These lines were added later, as Anthropic's own pages caught up, and they sit beside the original rather than replacing it.
Added since
A small documentation edit on Run Claude Code programmatically touched a line naming MCP_TIMEOUT after this was published.
Use the MCP server fields the same way. When you pass [`--mcp-config`](/docs/en/cli-reference#cli-flags) with `-p`, Claude Code waits for still-pending servers before running the first turn, up to the [`MCP_TIMEOUT`](/docs/en/env-vars) sta…headless see the edit
Confirmed since
Anthropic's documentation has since written up MCP_TIMEOUT, on Error reference.
The tool you passed to [`--permission-prompt-tool`](/docs/en/cli-reference#cli-flags) wasn't among the connected MCP tools when the run first needed a permission decision, either because its server never connected or because no connected s…errors see the edit
Two sources agreeTwo things we can check say the same as this entry.
Anthropic's documentation agrees
Anthropic's documentation has since written up MCP_TIMEOUT, on Error reference.
Anthropic's release notes agree
Improved screen-reader output in /mcp: a disabled server is read as "off" instead of "pending"