MCP tools
A failed tool call returns a normal MCP result withisError: true. Its one text block says what failed, the status behind it, what to do next, and a reference id:
Reference id when you ask Rippit about a failure. It identifies the exact request.
Arguments that don’t match a tool’s schema are rejected before the tool runs, with text that starts
Input validation error and describes each problem.
When a retry is safe
Reads and any call that returned429 are safe to retry. A write that timed out or returned 5xx may have taken effect, so check before repeating it. Idempotency and retries lists which operations are safe to repeat and how to check.
Connection errors
Before any tool runs, the server checks the request’s sign-in. These failures are plain HTTP responses with a JSON body and anx-request-id header:
Bulk API
Errors return a JSON body with adetail field. For most statuses it is a message:
422 it is a list with one entry per invalid field, where loc is the path to the field:
Job errors
A request can succeed while the job it started does not. Read the job untilstatus is done or failed:
- A
failedjob has the reason inerror. - A
donejob can still have skipped rows. Each one is inresult.rejectedwith its 0-basedrownumber and the reason, and every other row was written.
Rippit CLI
A failed command exits with status2 when the command line itself was rejected, and 1 for every other failure. With --format json, stderr holds one object:
retryable is false, check the result before running the command again. See Rippit CLI for the codes each command returns.