GraphIngest

GraphIngest runs your workflows. This listing lets an agent watch them and fix them.

The daily reality of orchestration is not writing pipelines, it is finding out that one broke. Something failed at 3am, and the loop is always the same: open the dashboard, find the run, read the logs, work out which task died, fix it, restart. Every step of that is mechanical, and none of it needs a human except the fix.

So the tools follow that loop rather than mirroring the API. list_flow_runs answers "did anything break" with a count by state and the failed run ids up front. get_run_details returns the run, every task, the tasks that actually failed, and the logs in one call - because an agent that needs three calls to explain a failure usually stops after the first and guesses. retry_run restarts from the point of failure, so the work that already succeeded is not repeated.

It authenticates with your own GraphIngest API key, entered once on your subscription page and never passed as a tool argument. That means your plan, your quotas and your revocation apply exactly as they do to an SDK call - the agent is another client holding your key, with no elevated access of any kind.

What it cannot do is as deliberate as what it can. There is no deploy tool, no key management, no billing, no environment variables. Deployment pushes code and secrets from your working directory and belongs on your machine; the rest is not work worth handing to an agent. The surface is read-and-operate, and stops there.

Full product at www.graphingest.io.