Braintrust

About

Know if your prompt change made things worse

Braintrust combines tracing, offline evals, online scoring, playgrounds and human review in one platform. The bt CLI instruments apps and runs evals from the terminal, and a remote MCP server connects Braintrust to Claude Code, Cursor and other clients. The free Starter plan includes unlimited users and projects.

Features

• Production tracing and online scores
• Evals on datasets and experiments
• Playgrounds for prompts and models
• bt CLI and remote MCP server

Install

• Claude Code$claude mcp add --transport http braintrust https://api.braintrust.dev/mcp
• Codex$codex mcp add braintrust --url https://api.braintrust.dev/mcp
• Cursor$npx -y mcp-remote https://api.braintrust.dev/mcp
• Grok Bot$grokbot mcp add braintrust https://api.braintrust.dev/mcp