[DOCS] MCP docs still describe sub-1000 ms per-server `timeout` values as 1-second flooring

Documentation Type

Incorrect/outdated documentation

Documentation Location

https://code.claude.com/docs/en/mcp

Section/Topic

"Managing your servers" tips block and the per-server timeout behavior paragraph

Current Documentation

The docs currently say:

# Get details for a specific server claude mcp get github

and later:

* Set a per-server tool execution timeout by adding a timeout field in milliseconds to that server's .mcp.json entry, for example "timeout": 600000 for ten minutes. This overrides the MCP_TOOL_TIMEOUT environment variable for that server only

The per-server timeout is a hard wall-clock limit per tool call, and progress notifications from the server do not extend it. Values below 1000 are floored to one second. For HTTP and SSE servers, the per-request fetch first-byte budget has a 60-second minimum regardless of this value, so only the tool-call watchdog honors smaller values.

What's Wrong or Missing?

As of v2.1.162, the release notes say Claude Code fixed the buggy behavior where per-server .mcp.json timeout values below 1000 ms were floored to a 1-second watchdog that aborted every tool call.

The current mcp page still documents that old behavior instead of the fixed behavior:

A. The per-server timeout rule is outdated

The page says values below 1000 are floored to one second, but v2.1.162 says sub-1000 ms per-server timeout values are now ignored and fall back to MCP_TOOL_TIMEOUT or the default timeout.

B. The claude mcp get verification path is incomplete

The page tells users to inspect server details with claude mcp get, but it does not mention that claude mcp get now annotates ignored sub-1000 ms per-server timeout values. That leaves readers without the documented way to confirm why a tiny configured timeout is not being applied.

Suggested Improvement

Update the mcp page to match v2.1.162:

Before:

Values below 1000 are floored to one second.

After:

Per-server .mcp.json timeout values below 1000 ms are ignored. Claude Code falls back to MCP_TOOL_TIMEOUT if it is set, otherwise it uses the default MCP tool timeout.

Then add a short note near the claude mcp get <name> guidance explaining that the command annotates ignored sub-1000 ms per-server timeout values so users can verify the effective behavior from the CLI.

Impact

Medium - Makes feature difficult to understand

Additional Context

Affected Pages:

| Page | Line(s) | Context |
|------|---------|---------|
| https://code.claude.com/docs/en/mcp | 143-150 | claude mcp get is presented as the way to inspect a server |
| https://code.claude.com/docs/en/mcp | 189-195 | Per-server timeout docs still describe sub-1000 values as 1-second flooring |

Total scope: 1 page affected

The relevant behavior change is called out in the Claude Code v2.1.162 changelog entry: sub-1000 ms per-server timeout values are ignored instead of being floored to a 1-second watchdog, and claude mcp get annotates that case.