Evaluates and observes LLM applications end to end: capture traces, score them with evals, iterate in a prompt playground, and gate deploys on eval results run from CI.
Use it now
Try this first
Ask your assistant, with Braintrust connected:
- Experiment Based Evals
Use Braintrust to run eval suites over datasets: ‹what you need›
- Production Trace Logging
Use Braintrust to capture traces: ‹what you need›
- Human Review And Datasets
Use Braintrust to collect human feedback on logged spans: ‹what you need›
Connect it over MCP
Connect Braintrust as an MCP server, following its setup instructions at https://braintrust.dev
Its page: braintrust.dev
Install the CLI
Install it as its own instructions say.
Its page: braintrust.dev
Call the API
Auth A bearer token from Braintrust Data
Its page: braintrust.dev
Compare
Press Compare to add it to the tray at the bottom of the window, then pick up to 2 others anywhere on the site.
- Access
- APICLIMCP
- Auth
- Bearer Token
- Skills
- 5
Skills
Experiment Based Evals
Runs eval suites over datasets with code and LLM-judge scorers, comparing scores across experiments and model versions.
Production Trace Logging
Captures traces and logs from production AI applications, then queries them to identify recurring failure patterns.
Prompt Playground
Iterates on prompts and models side by side, running them against dataset rows before the change ships to production.
Human Review And Datasets
Collects human feedback on logged spans and promotes interesting production examples into evaluation datasets.
Braintrust CLI
Runs evals, instruments code, queries logs, syncs data, and manages functions from the bt command line.
