Two Places to Look

Every action an AI agent takes in BFF is logged and visible. There are two views:

ViewWhereWho can see it
Agent ActivitySidebar → Team group → Agent Activity (/dashboard/agent-logs)Company admins only
Execution HistoryThe agent’s detail view, per agentAnyone who can open the agent

The Agent Activity Page (Fleet-Wide)

Agent Activity is the admin-only command center: “Execution history across all AI agents in this company.” Use it to answer “what did our agents do this week?” without opening agents one by one.

The table shows one row per execution:

  • Status — completed, partial, failed, or in progress (plus a dry badge for dry runs)
  • Agent and Task — who ran what
  • Steps — progress as completed/total (e.g. 4/5)
  • Duration and When

You can search and filter by status. Click any row to open the Execution Detail view.

Execution Detail

The detail view shows the full picture of a single run:

  • Status, agent, task, steps, duration, and when it ran
  • How the run was triggered (e.g. via webhook)
  • A Summary of what the agent did
  • Steps — each step’s number, action, output, and any decision taken

Document Drafting in the Logs

Agents with Enable Document Write show up here too. When an agent drafts or proposes a Policy, SOP, Work Instruction, or Training, that activity is logged alongside task executions — so admins can audit exactly what an agent wrote and when, before a human reviews and publishes it. See AI Agents as Document Authors for how proposals reach a human approver.

Per-Agent Execution History

For a single agent, open its detail view from AI Agents and scroll to the Execution History section — the same execution records, scoped to that agent. This is the quickest path when you’re already configuring or troubleshooting one agent.

Using Logs for Troubleshooting

When a run fails:

  1. Find the run in Agent Activity (filter by failed)
  2. Open the Execution Detail and locate the failed step
  3. Review the step’s output and decision
  4. Check whether the linked Work Instruction needs updating — then fix and re-run

Tip: Watch for runs marked partial — the agent finished some steps but not all. A recurring partial on the same step usually means the Work Instruction is ambiguous at that point, or the agent needs clearer Agent Instructions.