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

# Quickstart

> Welcome! Run your first analysis in Docent

**This quickstart guide will help you run your first Docent analysis on our sample Terminal-Bench data in a few minutes.** By the end, you'll understand how to run several common workflows, including analyzing common agent failures and diagnosing a model regression.

## Before you begin

1. Create an account at [docent.transluce.org](https://docent.transluce.org/signup) and sign in
2. Install the Docent SDK. We recommend using `uv`.

```shell theme={null}
cd your-project
uv init && uv add docent-python
```

3. Ensure you have [Claude Code](https://claude.com/claude-code) installed. Start it in your project with `claude`

## Instructions

<Steps>
  <Step title="Install the Docent plugin">
    <Tabs>
      <Tab title="Claude Code (Recommended)">
        Add the Transluce marketplace, then install the Docent plugin:

        ```shell theme={null}
        claude plugin marketplace add TransluceAI/claude-code-plugins
        claude plugin install docent@transluce-plugins
        ```

        Restart Claude Code after installation. Type `/` inside your Claude Code session to verify that `/docent` is available.

        <sub>*To update later, run `claude plugin update docent@transluce-plugins`.*</sub>
      </Tab>

      <Tab title="Cursor/VSCode">
        Install [uv](https://docs.astral.sh/uv/getting-started/installation/) first. The Docent MCP server in the template is launched through `uv`.

        Download and open a ready-made template in its own IDE window so the workspace configuration is picked up correctly.

        * [Download Cursor template](https://docent-public-assets.s3.us-east-1.amazonaws.com/cursor-quickstart.zip)
        * [Download VS Code template](https://docent-public-assets.s3.us-east-1.amazonaws.com/vscode-quickstart.zip)

        The template includes:

        * `pyproject.toml` with `docent-python`
        * `docent.env` for `DOCENT_API_KEY`
        * Agent instructions in `.cursor/rules/docent.mdc` or `AGENTS.md`
        * MCP configuration in `.cursor/mcp.json` or `.vscode/mcp.json`

        If you want to use Docent in an existing Cursor or VS Code workspace instead, copy the relevant template files into your project:

        * `pyproject.toml` dependency on `docent-python`
        * `docent.env`
        * `.cursor/rules/docent.mdc` or `AGENTS.md`
        * `.cursor/mcp.json` or `.vscode/mcp.json`
      </Tab>

      <Tab title="Other IDEs">
        Download the skill files directly and pass the file contents directly in your prompt context:

        * [Docent SKILL.md](https://github.com/TransluceAI/claude-code-plugins/blob/main/plugins/docent/skills/docent/SKILL.md)
      </Tab>
    </Tabs>
  </Step>

  <Step title="Add your API key to `docent.env`">
    Create a `docent.env` file in your workspace:

    ```shell theme={null}
    cat <<'EOF' > docent.env
    DOCENT_API_KEY=<YOUR_API_KEY>
    DOCENT_DOMAIN=docent.transluce.org
    EOF
    ```

    Then generate a Docent API key and paste it into that file:

    1. Create a new API key in [Settings](https://docent.transluce.org/settings/api-keys).
    2. Copy the key into `docent.env` as the value of `DOCENT_API_KEY`.
  </Step>

  <Step title="Run your first analysis">
    The prompts below contain the collection ID of our sample Terminal-Bench collection. To use your own collection, copy the ID in the top left corner of your collection, next to the collection name.

    <Tabs>
      <Tab title="Failure modes">
        ```text wrap theme={null}
        /docent What are the main reasons why GPT-5.1 Codex fails?

        Identify runs where GPT-5.1 failed. Summarize the primary failure modes in those runs and explain why you think they were decisive. Cluster common failure modes or failing strategies across all runs. Continue to cluster within clusters until you reach failures that are prevalent (i.e. common in the data) and specific (i.e. it is evident to a developer what a concrete fix would look like).

        - Collection ID: 479b7093-5a33-47f1-8d7b-fc9f6f16bb75
        - Auto accept reading plan and generate a report
        ```
      </Tab>

      <Tab title="Compare models">
        ```text wrap theme={null}
        /docent What are the main reasons why GPT-5.1 Codex underperforms GPT-5 Codex?

        Identify tasks where GPT-5.1 regresses on average. On those tasks, compare a failed GPT-5.1 run against the successful GPT-5 runs. Summarize the main failure modes and analyze whether avoiding those failures was material to the result of the successful runs.

        - Collection ID: 479b7093-5a33-47f1-8d7b-fc9f6f16bb75
        - Auto accept reading plan and generate a report
        ```
      </Tab>

      <Tab title="Collection Overview">
        ```text wrap theme={null}
        /docent Give me an overview of this collection. Generate a report.

        - Collection ID: 479b7093-5a33-47f1-8d7b-fc9f6f16bb75
        - Auto accept reading plan
        ```
      </Tab>
    </Tabs>

    Claude Code will generate a script using the Docent SDK to orchestrate your analysis. When it finishes, you'll be able to view the results in the Docent UI.
  </Step>
</Steps>

## Next steps

<Card title="Ingest your own data" icon="upload" href="/ingestion/overview" horizontal>
  Ready to analyze your own agent runs? Follow the ingestion guide to load your logs into Docent.
</Card>

* Any questions or difficulty ingesting your data? Chat with us directly via [Slack](https://transluce.org/docent/slack).
* Handling sensitive data? [Reach out to our enterprise team](mailto:docent@transluce.org).