Every time an agent is asked something, Pawabase keeps a run: the question, each step, every tool used, the answer and what it cost.

​
Where to find runs

  • AI → Agents → an agent → Runs lists everything that agent has been asked, from Studio, your app or a flow.
  • AI → Agents → Runs lists runs across all agents.
Filter by agent, by status (running, waiting, done, failed) or by conversation. Open a run to read it.

​
What a run shows

Someone running an agent in your app sees what happened but not the raw inputs and outputs of each tool; only your servers see those.

​
Conversations

Runs that share a thread are one conversation. The agent remembers recent messages in a thread, up to the last 60. See Run an agent from your app.

​
Start a flow when a run ends

Pawabase announces what happens to a run as events, so a flow can react: Typical uses:
  • Send a Slack-style notice to a manager when a run is waiting for approval.
  • Save the answer of a long background run to a table when it finishes.
  • Email a summary when a scheduled agent run ends.
Create an event subscription for the event and pick the flow to run.

​
Clean up

Delete a run from its page when you no longer need the record.

​
Check usage and cost

AI → Usage shows calls, tokens and cost for the models your agents use. Set a Most cost limit on each agent so one run cannot surprise you.