Braintrust
EvaluationApache-2.0 (SDK repositories)
Braintrust is a hosted platform for tracing agent runs, running evals on datasets and improving prompts and scorers. Developers instrument applications with the TypeScript or Python SDK or the bt command-line tool, and coding agents can connect through a hosted MCP server to query logs, run evals and edit prompts. The SDKs are open source under Apache-2.0. The pricing page lists a free Starter plan, a monthly Pro plan and Enterprise terms that include on-prem or hosted deployment. The home page says it is built for teams running agents in production.
The JavaScript SDK repository was previously named braintrust-sdk and is now braintrust-sdk-javascript; the Python code lives in braintrust-sdk-python.

Pinout
- MCP server: supported, verified
- HTTP API: not checked yet
- command line: supported, verified
- Python SDK: supported, verified
- TypeScript SDK: supported, verified
- OpenAPI description: not checked yet
| Interface | Status | Source we read | Read on |
|---|---|---|---|
| MCP server | Supported, verified | Braintrust MCP | |
| HTTP API | Not checked yet | — | — |
| Command line | Supported, verified | CLI reference | |
| Python SDK | Supported, verified | Python SDK repository, README | |
| TypeScript SDK | Supported, verified | Repository home page, README | |
| OpenAPI description | Not checked yet | — | — |
Facts
| Pricing, starter plan | $0 per month; $10 per month of model credits, 1 GB processed data, 10K scores, 14-day retention; the page states no credit card is required. Pricing page, read . |
|---|---|
| Pro plan | $249 per month; $100 per month of model credits, 5 GB processed data (then $3 per GB), 50K scores (then $1.50 per 1,000), 30-day retention. Pricing page, read . |
| Enterprise deployment | Custom terms with on-prem or hosted deployment. Pricing page, read . |
| Hosted MCP server | US https://api.braintrust.dev/mcp, EU https://api-eu.braintrust.dev/mcp; streamable HTTP; SSE-only clients cannot connect. Braintrust MCP, read . |
| MCP authentication | OAuth 2.0 (recommended) or an API key sent as a bearer token. Braintrust MCP, read . |
| MCP tools | 52 tools listed, including sql_query, run_eval, create_evaluator and search_docs. Braintrust MCP, read . |
| MCP clients named | Claude Code, Cursor, Codex, Gemini CLI, GitHub Copilot CLI, VS Code and Zed, among 19 listed. Braintrust MCP, read . |
| CLI | bt command-line tool: curl -fsSL https://bt.dev/cli/install.sh | bash, or npm install -g @braintrust/bt; includes bt eval and bt sql. CLI reference, read . |
| TypeScript SDK | npm install braintrust autoevals; package braintrust covers logging, tracing, evals and the CLI. Repository home page, README, read . |
| Python SDK | pip install braintrust autoevals (separate repository braintrust-sdk-python). Python SDK repository, README, read . |
| SDK license | Apache-2.0 for both the JavaScript and Python SDK repositories. Repository home page, README, read . |
| Other integration packages | @braintrust/otel (OpenTelemetry span processor), @braintrust/temporal, @braintrust/browser. Repository home page, README, read . |
| Native SDK languages named | Python, TypeScript, Go, Ruby and C#, among others. Home page, read . |
| Last commit | . GitHub API, latest commit on the default branch of braintrustdata/braintrust-sdk-javascript, read . |
| Frameworks | OpenTelemetry (Repository home page, README, read 2026-11-16); Temporal (Repository home page, README, read 2026-11-16); Claude Code (Braintrust MCP, read 2026-11-16); Cursor (Braintrust MCP, read 2026-11-16). |
Limits
- The MCP server needs streamable HTTP; SSE-only clients cannot connect.
- On-prem or hosted deployment is listed under Enterprise.
Strengths
- Free Starter plan with no credit card required.
- MCP server supports OAuth sign-in or an API key.
- SDKs in TypeScript and Python are Apache-2.0 licensed.
Pricing
As written on Braintrust's own page on the read date. Prices change: check there before you pay.
Alternatives to Braintrust
All evaluation tools for agents
Open-core platform to trace, test and evaluate agents and coding assistants.

Framework for evaluating, testing and monitoring ML and LLM systems.

Command line tool and library to evaluate and red-team LLM apps and agents.

Python framework for testing LLM outputs and agents with metrics.
Last check: . Method: public repository and docs read automatically, page and day written down. Next check due within 90 days. How we check
Is this your tool? Write to hello@agent-tools.directory from an address on its domain to correct a pin or a fact.