Skip to main content
Claude Cowork extends Claude’s agentic capabilities to general knowledge work, using the same architecture as Claude Code without a terminal. It runs in the Claude desktop app, on the web, and on mobile. Braintrust provides the same two plugins for Cowork that it provides for Claude Code:
  • braintrust: Brings context (docs, logs, experiments) from Braintrust into Cowork. For example, query logs, access Braintrust docs, or fetch experiment results.
  • trace-claude-code: Traces Cowork’s operations to show LLM calls, tool usage, and timing data. This can be useful for personal exploration or to monitor your team’s activity.
Cowork installs plugins from the same marketplace as Claude Code (braintrustdata/braintrust-claude-plugin). This page covers installing them in Cowork. For full plugin usage, configuration, and troubleshooting, see the Claude Code guide.

​
Setup

1

Open Cowork

In the Claude desktop app or on claude.ai, select Cowork in the message box.
2

Add the Braintrust plugin marketplace

In the sidebar, click Customize, go to Plugins, then click Add > Add marketplace and enter:
3

Install the plugins

In Customize > Plugins, install braintrust, trace-claude-code, or both from the marketplace.
4

Set your API key

Both plugins authenticate using BRAINTRUST_API_KEY. Set it as described in the Claude Code prerequisites. Cowork shares its configuration with the Claude desktop app.

​
Usage

Once installed, the plugins work the same way in Cowork as in Claude Code:
  • Use the braintrust plugin to query experiments, logs, and datasets directly in Cowork conversations. See Plugin: braintrust.
  • The trace-claude-code plugin automatically traces each Cowork session to Braintrust as a hierarchy of spans. See Plugin: trace-claude-code.
To view your traces, go to your project in the Braintrust UI and select Logs.

​
Next steps

  • Configure the plugins: See the Claude Code guide for full plugin usage, configuration, and troubleshooting.
  • Run evaluations: Check out the evaluation guide to learn evaluation patterns.
  • Explore MCP tools: See the MCP documentation for all available commands.