Skip to main content
An autonomous agent that acts on your tools without you watching every step only works if you can trust what it did. The activity timeline is how: a real-time, per-agent record of every turn, tool call, and decision, always available and never something you have to ask the agent to summarize for you.

​
The timeline

Open an agent’s page and expand the Activity panel in the side panel. Events stream in live as the agent works — nothing to refresh.

​
Anatomy of an event

Each event carries: Typical events read like a log of a colleague’s work: visiting a page, reading or writing a file, updating memory, running code, calling an integration, starting a sub-agent, or replying to you. Grouping by origin makes it easy to answer “why did the agent do this?” — a badge tells you whether it was you asking in chat, an incoming email, a webhook firing, a scheduled cron run, another agent delegating work, or the system itself.

​
Sub-agents

Delegated work shows up two ways: as start/finish events inline in the timeline, and as their own dedicated Sub-agents section showing which are still in progress and which have finished, each with a live status.

​
Per-agent integration log

The Tool logs panel on each agent’s page lists every call it made to a connected integration — the tool used, and whether it succeeded, was denied, or errored. This is the fastest way to check what a specific agent has actually done with the access you granted it.

​
Company-wide view (admins)

Admins get a broader lens across every agent, from the company dashboard:
  • Integration logs — every tool call across your whole company, filterable by team, agent, and date range, with CSV export for both the filtered view and the full log.
  • Integrations usage — call volume aggregated per integration provider, so you can see at a glance which tools your agents actually rely on.
The company-wide view is where to look when auditing access after the fact — for compliance, for investigating an unexpected action, or just to understand usage patterns across the team.

​
Why this matters

Every action an agent takes through a granted integration is logged — which agent, which action, when, and with what outcome. Because agents can act unattended across chat, email, webhooks, cron, and other agents, the activity timeline and integration logs are what let you extend real autonomy to an agent without losing visibility into what it’s doing.

Chat

Where the Activity panel lives day to day.

Permissions

How grants control what an agent can call.

Agents

Sub-agents and how delegation works.

Credits

Where model usage cost shows up per turn.