# Studio (/docs/studio)

> Inspect runs, approve tool calls, and chat with your agent.

Studio is a local development tool. Run it against a project on your own machine, not deployed
over production data, because it can approve tool calls and cancel runs.

Start the bundled Studio with `plico serve --studio <project>`. Beside a backend that embeds Plico, add `--inspect` so Studio only reads; see [Use Plico from your backend](/docs/backend).

Studio has seven views in its left rail: Chat, Overview, Runs (the timeline, raw events and
replay), Instructions, Artifacts, Evals and Readiness.

## Approvals

Studio shows the run's event history and, when a tool call needs approval, its
reason and input arguments. Approve the call to resume the run or deny it to stop
the run before the tool executes. Each decision is recorded in the run's event log. See
[Approvals](/docs/approvals) for the policy and the other ways to decide.
The run inspector also lets you cancel active runs.

## Chat

In Chat, replies stream in as Markdown, **Stop** cancels the reply in progress, and
**Inspect run** opens the reply's timeline, where streamed model chunks are grouped
into one row. The thread header shows the tokens used by all its runs, and the run inspector shows
the tokens for one run. Configure a model provider for the project before sending messages.
See [Threads](/docs/threads).

## Readiness

The Readiness view reports approval policy overrides that reference missing tools or
conflict with a tool's declared approval requirement. It also reports whether
model-facing artifact tools are configured as `read-write`, `read-only`, or `off`.

## Artifacts

Open a run's **Artifacts** view to preview files, compare stored versions, and
download the current workspace file. See [Artifacts](/docs/artifacts).