What each error from the G2X MCP server means, what to do next, and how limits work.
| Status | Meaning | What to do |
|---|---|---|
401 | Not signed in, or the sign-in ended | Refresh the token, or sign in again |
403 with not_entitled | The account's plan doesn't include this tool | Nothing to wait for; the call won't succeed on this plan |
405 | A GET to an MCP endpoint | Send JSON-RPC as POST |
429 | A limit was reached | Wait for the time the message names, then try again |
503 | G2X is busy and the request didn't run | Try again after a few seconds |
Each plan sets a pace, a number of requests a minute shared by all of an account's connections, and how much the account can use. Checking usage or an analysis run's status doesn't count toward the pace.
A request past a limit answers 429 with a message that names the limit and when to try again, for example “You've reached your plan's limit of 30 requests a minute. Try again in 2 seconds.” Wait that long: a request sent sooner is refused the same way. Don't split one question into many calls to get around a limit.
A tool that can't complete a call returns isError: true and a message. Correct the input from the tool's schema, or take the step the message names; repeating the same call gives the same answer. When no record matches, refine the question rather than retrying.
Retry only 429, after the wait it names, and 503, after a few seconds, with a bounded number of attempts. For an analysis run, keep its run ID and check its status before starting another.