Skip to main content
Logs is the record of every request your agents made through Unbound — the prompt, the model, the tokens and the outcome.

What’s on the page

Tabs

Both tabs need the same Logs · View as the page.

All Logs

Tasks

Filters and controls

Both tabs carry a search box, faceted filters and a time range. All Logs filters by tool type, user, model and status; Tasks filters by tool type, user and model. The time range defaults to the last 24 hours, takes a custom range, and is remembered between visits. In All Logs the search also matches a trace ID. In Tasks the search covers user and model, and begins at three characters. Each tab also has a column picker for hiding columns you do not need, and columns can be dragged wider.

What you can do

Task trace

A task’s trace lists its turns in order with the total cost. Select a turn to see what was asked, what the agent did and what it replied. The page is reached from Logs → Tasks (breadcrumb Logs › Tasks › Task Details) and is built from these parts: Each turn row shows its number, the prompt, a type · outcome chip where a policy acted (for example Terminal · Block), the time, the model, the cost (or Cost unavailable), and its status code — a refused turn reads 446 with a warning mark, a normal one a tick. A turn a policy blocked is highlighted. Selecting a row opens it into User Request, Tool Actions and Assistant Response — the reply text only, with the tool calls listed under Tool Actions. A task link can end in ?turn=N to open the task on the page holding interaction N, with that turn expanded. Cloud agents. A turn run by a cloud agent has no signed-in user. Where the GitHub integration is connected, the row names the GitHub user who ran the session, with Cloud agent beneath; otherwise it reads as Cloud Agent with the repository it ran in, or No signed-in user.

Common questions

How do I export logs? Export from the table controls - the CSV covers your current filters and time range. Exporting here needs Analytics · View and Logs · View, with Organization scope. Where do I see what an agent actually did, rather than what it was asked? Analytics — Agentic Use, which classifies each command and tool call and matches it against your policies. How do I see a whole task from start to finish? The Tasks tab, then select a task for its turn-by-turn trace.