Run history

Every run leaves a clear trace you can read back and rerun

Who launched it, what the agent did step by step, what it produced and what it cost. You check its work without opening a log file.

All runs

Every run in a space, in one place

Each time an agent is launched, it becomes a run: by hand, by a schedule, through the API or from your own application. The list shows its status and what triggered it.

  • Filter by kind and by status: success, failed, running, pending, timeout, cancelled.
  • The "Trigger" column names the schedule, person, API key or tool that launched the run; the kind filter tells Manual, Schedule, API, End user and CLI apart.
  • Administrators and builders see every run in the space, and "My runs" narrows it to theirs. Other roles see only their own.

Results

What the agent produced, ready to use

The "Results" tab gathers the work handed back: files, structured data and what the agent kept in mind for next time.

  • "Productions": the files created during the run, with a preview.
  • "Structured output": the fields defined for the agent, read as they are or as JSON.
  • "Memory": what the agent noted during the run, to use in later runs.

What it changes

For your teams

  • Check what the agent did before you rely on its result.
  • Find the file it produced without digging through emails or folders.
  • Understand a failure and rerun, without waiting for the tech team.

For the organization

  • Every run says who or what launched it, with which agent version and which connections.
  • Space leads see every run in their space; everyone else sees their own.
  • A run heading the wrong way stops in one click.

When it breaks

A failure explained in plain words, then rerun

When a run fails, the "Overview" says why in everyday language and points to the step of the journal involved. The context is right there: who launched it, with which version, which connections.

  • A refused model key or a rate limit shows in plain words, with what to do next.
  • "Re-run" starts again with the same inputs; "Cancel" stops a running run.
  • The "Execution", "Activity", "Results" and "Usage" cards sum the run up: version, who launched it, number of tool calls and errors, cost.

Frequently asked questions

How long are a run's details kept?
Details have a retention period: after it, the run shows "Details expired".
Who sees other people's runs?
Administrators and builders see every run in the space. Other roles see only their own. A custom role can be given the "View every run in the space" permission.
Can I export or delete runs?
On our roadmap: export, per-run delete and a date filter are not in the interface yet.
How do I find a step in a long journal?
Search the journal, then filter by level (Information, Warning, Error) and by type (Message, Tool call, Tool result, Diagnostic). "Hide tools", "Timestamps" and "Copy logs" round it out.
How long can a run last?
Up to 30 minutes, with 1.5 GB of memory and 2 vCPUs. "Cancel" stops it earlier.
Is there a log of permission changes?
Changes to permissions, members, keys and connections are recorded in the database, append-only, with their author. Viewing them in the interface and through the API is on our roadmap.

Let's see what it would look like on your side.

Try it on your own in a few minutes, or take thirty minutes with us to scope a first use case.

To try it

Try it for free

Create a cloud account and describe your first agent.

For a guided start

Let's talk about your case

Thirty minutes with us to scope a first use case and its gain.

To audit

Read the code

Apache 2.0. Your technical team can read it all, install it all, keep it all.