05 Governance

Run history.

Every turn that uses tools leaves a trace.

Open the steps behind an agent's reply, look back over every Automation run, and see who acted on each card in the thread.

Where this sits in the harness iceberg
  • Run history
Where it sits. Each dot is another part of the harness; select one to open it.
How it works

The record you work with.

01

A trace behind the reply

When an agent uses tools, its reply carries the steps it took, shown as Worked for and the time it took. People with access to that agent can open it.

02

Private to the agent's audience

Traces follow the agent's access list, so workspace admins do not see them by default.

03

Automation runs

Each Automation keeps its runs with outcome, duration, outputs and credits, grouped by day, with a filter for failures.

04

Receipts in the thread

Rooms note who added an agent, and draft and approval cards show who acted on them.

On the surface

A Worked for line under an agent's reply that opens into its steps.

Without it

When something goes wrong, nobody can say who asked for it or what the agent did.

With Kylon

The steps, the runs and the decisions sit next to the conversation they belong to.

Questions

Good
questions.

Is this an audit log?

No. It is the record you work with: traces, run history and receipts in the thread. There is no separate audit log export.

Can I see what a run cost?

Automation runs show their credits. Usage by room, agent and person is in billing.