Braintrust
Braintrust is an MCP server that lets agents query AI experiments, logs, datasets, and prompts on Braintrust.
About
Braintrust MCP connects your coding agent to Braintrust’s evaluation and observability platform so you can inspect experiments, production-style logs, datasets, and prompt versions without leaving Claude Code or Cursor. developers shipping agent features need a loop between code changes and measurable quality; this server exposes that loop over streamable HTTP at Braintrust’s API host, with an optional EU data plane suffix when residency matters. Use it in Ship when you are running evals, comparing prompt variants, or digging into traces before you call a release done. It complements unit tests for deterministic code but targets non-deterministic LLM behavior. You will need a Braintrust account and API access configured in your client. The integration is read/query oriented from the catalog description—plan workflows around fetching eval artifacts and logs, not replacing your CI runner.
- Remote streamable-http MCP at api.braintrust.dev with optional EU region (-eu)
- Query experiments, logs, datasets, and prompts from the agent
- AI evaluation and observability platform surfaced over MCP
- Registry version 1.0.0; data plane region variable in manifest
- Supports pre-ship prompt and trace review for solo AI features
Braintrust by the numbers
- Data as of Jul 7, 2026 (Skillselion catalog sync)
claude mcp add --transport http braintrust https://api{region}.braintrust.dev/mcpAdd your badge
Show developers this MCP server is listed on Skillselion. Paste this into your README.
| Transport | HTTP |
|---|---|
| Auth | None |
What it does
Query Braintrust experiments, logs, datasets, and prompts from your agent while you eval prompts and traces before release.
Who is it for?
Best when you're shipping LLM features and already use Braintrust and want eval and observability queries inside the IDE agent.
Skip if: Greenfield projects with no eval harness and no need to compare prompts or trace production-like runs.
What you get
After registering the Braintrust MCP remote, your agent can pull experiments, logs, datasets, and prompts into the same thread as your code changes.
- Agent-accessible queries against Braintrust experiments and logs
- Faster Ship-phase review of datasets and prompts alongside code
By the numbers
- Registry version 1.0.0
- Remote URL pattern https://api{region}.braintrust.dev/mcp with region '' (US) or '-eu'
- Streamable-http transport type in server manifest
Recommended MCP Servers
Agent Audit McpImmutable audit trail for agent-to-agent interactions2
Agentguard47Read-only MCP server for coding-agent traces, alerts, costs, usage, and budget health.4
Agentic Observability MCPAgent tracing, cost tracking, anomaly detection for LLM agentsAgent Incident Relay McpAgent Incident Relay MCP — Article 73 5-clock broadcaster.
Agent Infra Mcp18 tools: agent memory, API drift detection, and HMAC-signed trust receipts
Agentmonitorrelay McpAI agent run monitoring with incident replay and SLA receipts.How it compares
Braintrust observability MCP bridge, not a generic logging agent skill.
FAQ
Who is Braintrust MCP for?
It is for developers and AI developers on Braintrust who want Claude Code, Cursor, or similar agents to access experiments, logs, datasets, and prompts over MCP.
When should I use Braintrust MCP?
Use it during Ship testing when you are validating prompts, reviewing traces, or comparing experiment results before release.
How do I add Braintrust to my agent?
Add the streamable-http remote https://api.braintrust.dev/mcp or https://api-eu.braintrust.dev/mcp for EU, with Braintrust API credentials per your client’s MCP setup.