- CollectionObservability & Tracing
- Classificationagent-native
- MCPAvailable
- Agent SkillUnknown
- InterfacesMCP · REST · SDK · http · API
What it does
Braintrust is an AI observability and evaluation platform: trace every LLM call, tool invocation, and intermediate step in production; score outputs with code, LLM judges, or humans; compare prompts and models; and convert traces to datasets for regression testing. It explicitly targets agents — e.g. native OpenAI Agents SDK trace processors — and exposes an MCP server so coding agents can query experiments and logs directly.
Agent-facing interfaces
- Catalog protocols: MCP, REST, SDK, http, API
- MCP: ✅ Hosted MCP endpoint for coding agents and IDEs.
- Agent Skill: ⚠️ No single npx skills add bundle verified; MCP exposes logs, evals, and prompt workflows to compatible hosts.
- Onboarding type:
sdk
How agents onboard
pip install "braintrust[openai-agents]"
See the full onboarding section in the dossier.
Protocol surface
| Interface | Detail |
|---|---|
| Python SDK | pip install braintrust / braintrust[openai-agents] |
| TypeScript SDK | @braintrust/* packages per docs |
| MCP | HTTPS api.braintrust.dev/mcp |
| REST API | https://www.braintrust.dev/docs/api-reference/introduction |