Observability - official Braintrust server
Braintrust MCP server
Hosted MCP server that lets AI tools read and write Braintrust data: query logs, experiments and datasets with SQL, author prompts and scorers, configure dashboards, alerts and Topics, and run evals. Write tools act with the permissions of the authenticated account. Braintrust hosts it at https://api.braintrust.dev/mcp with OAuth or API key auth.
Last verified 15 tools2 sources
Install
Cursor
About Cursor{
"mcpServers": {
"braintrust": {
"url": "https://api.braintrust.dev/mcp"
}
}
}Tools
sql_query- Query experiments, datasets, and logs using SQL.
infer_schema- Discover available fields, data types, and most common values in experiments, datasets, or logs.
summarize_experiment- Get aggregated performance metrics for an experiment, optionally compared to a baseline.
list_recent_objects- List recently created projects, experiments, datasets, prompts, or functions you have access to.
resolve_object- Convert names to IDs or vice versa, and parse Braintrust URLs.
generate_permalink- Generate a direct web link to a Braintrust object for sharing or bookmarking.
get_trace_work_items- Return meaningful LLM, tool, function, and task spans of a trace in chronological order.
search_patterns- Search existing project patterns by text, semantic similarity, ID, status, or quality.
generate_monitor_chart- Preview a monitor chart for project logs without modifying a saved view.
create_monitoring_view- Create a project-scoped monitor view, optionally containing previewed charts.
create_threshold_alert- Create an alert for an aggregate over a recent window of project data, evaluated on a schedule.
create_prompt- Create a versioned prompt from a completion-style prompt or chat messages.
create_evaluator- Create a versioned LLM or inline code evaluator.
run_eval- Run an experiment with a hosted dataset, inline rows, or a prior experiment as input, plus saved or inline scorers.
search_docs- Search Braintrust documentation for guides, API references, and code examples.
As documented on October 17, 2026. A running server reports its current list through tools/list.
Notes and caveats
- Remote only, nothing to install.
- EU data-plane endpoint is https://api-eu.braintrust.dev/mcp; self-hosted deployments use the MCP URL shown in Settings > Data plane.
- Streamable HTTP only; SSE-only clients cannot connect.
- OAuth uses dynamic client registration; API keys are sent as a bearer token.
- The earlier braintrust plugins for Claude Code and Codex are retired in favour of native MCP configuration.
- Braintrust docs give no verbatim config for Claude Code, Codex or VS Code; they link to each client's own MCP guide and give only the endpoint and auth header.
- The Cursor snippet is the docs' OAuth manual config for ~/.cursor/mcp.json (sign in by asking the agent, or
agent mcp login braintrustin the Cursor CLI); the docs' API-key variant adds "headers": { "Authorization": "Bearer ${env:BRAINTRUST_API_KEY}" }. - A Braintrust plugin for Cursor (US data plane, API key) can also add the server.
- Docs list 40+ tools (15 shown). run_eval and Topics rewinds incur compute/model usage.
- Docs carry no beta or preview label and no GA statement.
More MCP servers for observability
- ArizeStatus: Preview
Arize AX MCP server
Hosted MCP server that gives agents access to Arize AX projects, traces, spans, datasets, experiments, evaluators, monitors and prompts. It is a thin layer over the Arize REST API v2.
Transport: Streamable HTTPAuth: OAuth or API key - ArizeStatus: GA
Phoenix MCP server
Local npm-distributed MCP server for Arize Phoenix that connects to a Phoenix instance with a base URL and API key. Covers projects, traces, spans, sessions, annotation configs, prompts, datasets and experiments.
Transport: stdioAuth: API key - ArizeStatus: Beta
Phoenix Remote MCP server
MCP server built into Arize Phoenix at the /mcp path of any Phoenix instance, letting agents read and act on traces, spans, datasets, experiments and prompts through Phoenix's REST API, plus read-only SQL and GraphQL queries. By default it exposes code-mode tools that search the operation catalog and run sandboxed Python that chains operations.
Transport: Streamable HTTPAuth: OAuth or API key - AWS
AWS Labs CloudWatch MCP server
Open source local MCP server from AWS Labs that gives troubleshooting agents access to CloudWatch metrics, alarms and Logs Insights for root cause analysis.
Transport: stdioAuth: Other - Axiom
Axiom MCP server
Hosted MCP server that lets AI agents query Axiom event, trace and metrics data with APL and MPL, explore datasets, and manage dashboards, monitors and notifiers.
Transport: Streamable HTTP, SSEAuth: OAuth or API key - Better Stack
Better Stack MCP server
Hosted MCP server that connects AI assistants to Better Stack Incidents and Telemetry: on-call, incidents, uptime monitors, status pages, ClickHouse queries over logs, metrics, traces and errors, dashboards, and alerts.
Transport: Streamable HTTPAuth: OAuth or API key
Related reading
- AI Agent Eval Tools Compared: Braintrust vs Promptfoo vs DeepEvalAI agent eval tools compared - Braintrust, Langfuse, Promptfoo, DeepEval, Ragas, and OpenAI Evals on offline vs online evals, LLM-as-judge, CI, and datasets.
- Langfuse vs Braintrust vs Helicone: Agent Tracing Tools in 2026Langfuse vs Braintrust vs Helicone, plus LangSmith and Arize Phoenix: 2026 pricing tiers, self-host options, agent-trace features, and OpenTelemetry support.
- What's New in Mastra (Late 2026): Harness, AgentController GA, Eval Gates and MCP 2.xMastra in late 2026: goals, schedules, signals, AgentController GA, eval gates and MCP 2.x, with type-checked TypeScript examples and honest tradeoffs.
- MCP vs A2A vs Function Calling: Which Layer Does What in 2026MCP vs A2A vs function calling compared on the 2026-07-28 MCP spec and A2A 1.0: what each protocol does, where they overlap, and how to stack them.
- Pydantic AI V2 in Late 2026: Capabilities, Graphs, Evals and MCPPydantic AI V2 explained through v2.57.0: capabilities, typed output, Conversation storage, caching, graphs, evals, MCP Apps and Logfire, with checked code.
- Reverse Engineer Anything (REA): What the MCP Server Does and Where the Line IsREA (Reverse Engineer Anything) is an MIT-licensed MCP server that lets Claude Code, Codex or Cursor inspect binaries, Electron apps and APKs. How to try it.
FAQ
What does the Braintrust MCP server do?
Hosted MCP server that lets AI tools read and write Braintrust data: query logs, experiments and datasets with SQL, author prompts and scorers, configure dashboards, alerts and Topics, and run evals. Write tools act with the permissions of the authenticated account. The vendor documents 15 tools, including sql_query, infer_schema, summarize_experiment and list_recent_objects.
How do I add the Braintrust MCP server to Cursor?
Braintrust documents a Cursor config; copy it from the Cursor section of this page.
Does the Braintrust MCP server need authentication?
Yes, OAuth or API key. Transport: Streamable HTTP. Status: Unlabelled, last checked October 17, 2026.
Sources
Last verified against 2 Braintrust pages, then re-checked independently.
Newsletter
New MCP servers, the week they ship
One email per week. New official MCP servers, the coding tools that support them and how developers wire them up.