openclaw - 💡(How to fix) Fix Support per-server requestTimeoutMs in MCP config [1 comments, 2 participants]

When using MCP servers that execute long-running operations (e.g., Redash queries against Redshift/Athena that take 1-3 minutes), the MCP request times out with error -32001: Request timed out.

TL;DR

Add a requestTimeoutMs property to the per-server MCP config to increase the timeout for long-running operations.

Guidance

• Identify the specific MCP server configurations that require a longer timeout and update their configurations with the proposed requestTimeoutMs property.
• Verify that the MCP SDK's Client.callTool() method is properly passing the timeout options from the config.
• Test the updated configuration with a long-running operation to ensure the timeout is correctly applied.
• Consider implementing the proposed change to pass the timeout from the config to the MCP SDK's Client.callTool() method.

Example

{
  "mcp": {
    "servers": {
      "redash": {
        "command": "npx",
        "args": ["-y", "@suthio/redash-mcp"],
        "requestTimeoutMs": 300000
      }
    }
  }
}

Notes

The current workaround using direct REST API calls via curl + async job polling can continue to be used until the proposed change is implemented.

Recommendation

Apply the proposed workaround of adding a requestTimeoutMs property to the per-server MCP config, as it directly addresses the issue of timeouts for long-running operations.