> ## Documentation Index
> Fetch the complete documentation index at: https://docs.phinite.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Studio evaluations

> Agent Evaluations panel — entry points and permissions.

Open **Evaluations** in the Graph Studio sidebar (`?tab=eval`). Panel title: **Agent Evaluations**. Header link **History** jumps to workspace **History / Sessions**.

<Frame caption="Agent Evaluations panel">
  <img src="https://mintcdn.com/phinite/1VctjSmXCA5rNwu2/images/v2/evals/03-studio-panel.png?fit=max&auto=format&n=1VctjSmXCA5rNwu2&q=85&s=e7feb605c32a8c8ec0f4bd2c5b6ed985" alt="Connect to Build, Run on draft, Experiment, Signal" width="1440" height="900" data-path="images/v2/evals/03-studio-panel.png" />
</Frame>

## Entry points

| Label                | Description                                             | Permission                           |
| -------------------- | ------------------------------------------------------- | ------------------------------------ |
| **Connect to Build** | Attach metrics to a build for live or scheduled scoring | `configure`                          |
| **Run on draft**     | Evaluate the current flow version with mixed metrics    | `run`                                |
| **Experiment**       | Compare builds or build vs draft                        | `configure` — **UI stub only** today |
| **Signal**           | Toggle signal ingest on a build                         | `configure`                          |

Copy under the title: *Choose how you want to evaluate this agent. Run history and analytics live on the workspace Evaluations page.*

## Signal panel

After **Signal**:

* Copy explains enabling/disabling signal ingest per build.
* Columns: **Build** · **Version** · **Since** · **Signal** (toggle).

## Experiment

Selecting **Experiment** shows placeholder copy (*Compare different builds, or build vs draft.*) without a full wizard yet. Prefer **Connect to Build** or **Run on draft** for production scoring.

## Deep dives

1. [Connect to Build](/evaluations/connect-to-build) — Mode → Build → Metrics → Review.
2. [Run on draft](/evaluations/run-on-draft) — Dataset → cases/sessions → Metrics → Launch.
3. [Workspace Analytics / History](/evaluations/overview) — results after runs complete.

## Related

* [Evaluations overview](/evaluations/overview)
* [Builds overview](/builds/overview)