Know what ran, and what broke

Every execute call, job, webhook and event run is kept with its logs, result and error. When one fails, Minion explains why and puts it in front of the person who can fix it.

Connect your agent What's waiting on you
A run list with one failed row leading to a bell and a person

Every run, on the record

Runs started by your agent and runs that happened while nobody was looking land in the same history.

Logs and results

Each run keeps its log lines, its result and how long it took, plus the full error and stack when it fails.

A diagnosis for each failure

Minion writes a note with the likely cause, the last good run, and what changed since: package versions, code, job settings, secrets, the upstream HTTP status. Repeats of one failure share a note.

Fix, re-run, close

Fix the package, re-run the job, and mark the failure resolved, or ignored if it's known noise. Its logs stay either way.

One list of what's waiting on you

Some things only a person can do. Minion keeps them in one place, each with a link that clears it.

Approvals

A new host for a secret, a webhook destination, or a remote MCP server whose tools changed, each waiting on an owner or admin.

Things about to go wrong

Secrets about to expire, integrations to reconnect, quarantined mail to release, and jobs that haven't had a green run yet.

Sent to the right person

Each item says who can clear it: you, any member, or owners and admins. Your agent sends the link instead of telling you to check a dashboard.

When a scheduled script fails quietly

Finding out
A cron job on a server: Someone notices the output stopped
A Minion job: An open error, and an item on the waiting list
Why it failed
A cron job on a server: SSH in and grep the logs
A Minion job: Logs, the error and a diagnosis note on the run
What changed
A cron job on a server: Guesswork
A Minion job: Versions, secrets and settings since the last good run
After the fix
A cron job on a server: Wait for tonight's run
A Minion job: Re-run it now and mark it resolved

Questions about runs

Can my agent fix a failed run?

It can read the logs and the diagnosis, then save a fix as an unpublished draft of the package for someone to review and publish.

What if my agent's request times out?

The run keeps going. Your agent can check the run until it finishes, or repeat the call to get the stored result.

Do runs count against my plan?

Execute calls and job runs have daily limits by plan, reset at 00:00 UTC. A run refused for quota shows up as an error, so you can see it. See pricing.

Let the job tell you when it breaks

Connect your agent, schedule something, and see every run it makes.

Create a free account Read the docs