Agents

An agent is an AI helper that decides which of your functions to run to get a job done. This view is about your agents running live in production: where they can be reached, and a full record of every time one was called and what it did. You can see your agents working, not just trust that they are.

Annotated screenshot of the deployed Agents panel showing a live agent endpoint URL and a table of recent invocations with timestamps and outcomes, callout: "Your agents, live and reachable — with a record of every time they were called."

Live and reachable

Each deployed agent has its own endpoint — a live address it can be called at. The panel shows that endpoint plus a history of recent invocations: when the agent was called, and how each call turned out.

• How often it runs. See the rhythm of invocations at a glance.
• What it did each time. Open an invocation to see the outcome of that call.

Whose identity does an agent act under?

Just like workflows, an agent acts as the person who triggered it. That identity flows through everything the agent does, so it can only take actions that person is allowed to take. Every invocation is written to the audit log, so there's always a record of what ran and on whose behalf.

An agent never acts as a faceless robot. Each run carries the identity of whoever triggered it, and the audit log records every invocation — so you always know what an agent did and for whom.

Is it misbehaving or running up cost?

Agents use AI, and AI usage has a cost — so it's worth keeping an eye on them. Two things help:

• What's powering it is your choice — Pikku Studio Cloud's built-in AI or your own provider key — and its spend is itemized on your Usage page. See AI providers.
• Its behavior shows up in your logs — what it called, how often, and how it went. See Logs to spot anything unusual or unexpectedly busy.

Design-time vs run-time

This is the running view. To inspect what an agent can do — the functions and tools it's allowed to use — before it runs, see the design-time view in Building agents.

Where to go next

• Building agents — inspect an agent's capabilities before it runs.
• Logs — watch agent behavior and catch anything unusual.
• Audit — the record of every invocation and who triggered it.