Braintrust Tracing FAQs

Question 1

Are there any local dependencies required?

Accepted Answer

Yes, you must have the Claude Code CLI installed along with the 'jq' command-line utility for processing JSON data within the hook scripts.

Question 2

How does Braintrust Tracing help with debugging?

Accepted Answer

It provides a hierarchical view of every action Claude takes, from the initial user prompt to specific file reads and terminal commands, making it easy to identify exactly where a workflow failed.

Question 3

What tools and commands are captured by the trace?

Accepted Answer

The skill captures every tool interaction, including file reads (read_file), file edits (edit_file), and terminal command executions, including their inputs and outputs.

Question 4

Can I use different Braintrust projects for different repositories?

Accepted Answer

Absolutely. The configuration is stored in the .claude/settings.local.json file of your project, allowing you to define a specific BRAINTRUST_CC_PROJECT for each workspace.

Question 5

Do I need a Braintrust account to use this skill?

Accepted Answer

Yes, you need an active Braintrust account and a valid API key to visualize the traces on the Braintrust dashboard.

Braintrust Tracing

About

Key Features

Use Cases

Braintrust Tracing

About

Key Features

Use Cases