
by Braintrust
Braintrust is an observability and evaluation platform for AI agents. It captures traces, surfaces patterns in them, and lets teams run evals to improve quality.
Web appFreemium
Step 1: Create a Braintrust account.
Step 2: Log your first trace by sending traces from your AI application.
Step 3: Run your first eval to compare prompts and models.
Step 4: Review patterns found in your traces, or ask Loop to investigate.
Step 5: Add human annotations and build datasets from real traces.
Step 6: Ship changes and monitor production with dashboards.
Limits cover processed data, scores, data retention and monthly model credits, and on-demand rates apply once you exceed the included amounts.
Yes. Built-in models are available on every plan and draw on monthly model credits. Calling them directly through the gateway needs Pro or Enterprise.
Enterprise offers flexible deployment options including bring-your-own-cloud, alongside the standard SaaS deployment.
The docs describe an MCP connection for AI tools and a bt CLI for instrumenting apps and running evals.
Braintrust AI ProxyUnified model proxy for routing AI application traffic across providers with logging, caching, retries, and evaluation workflows
Braintrust EvalsBraintrust Evals is an AI-native product for evaluation SDK and platform for testing prompts, models, scorers, datasets, and AI application regressions.
PopSQL AIPopSQL AI uses AI for collaborative SQL editor with AI assistance for generating, explaining, and fixing queries against connected databases.
Hex MagicHex Magic uses AI for AI analytics assistant for generating SQL and Python, exploring data, debugging analysis, and building data applications.