Braintrust

Braintrust (braintrust.dev) is an end-to-end platform for building, evaluating, and observing AI applications. Its REST API at api.braintrust.dev exposes projects, experiments, datasets, logs/spans, prompts, functions and scorers, evals, and full organization/ACL management, plus an OpenAI-compatible AI proxy, all authenticated with a Bearer API key.

10 APIs 0 Features

AILLMEvaluationObservabilityLLMOps

APIs

Braintrust Projects API

Create, list, update, and delete projects - the top-level containers that organize AI features, experiments, datasets, logs, prompts, and scorers.

Braintrust Experiments API

Manage evaluation experiments and their events - create experiments, insert and fetch events, submit feedback, summarize results, and launch evals.

Braintrust Datasets API

Manage datasets and dataset events used as test cases - create datasets, insert, fetch, and summarize records, and submit feedback, plus dataset snapshots.

Braintrust Logs and Spans API

Insert, fetch, and provide feedback on production trace spans and log events stored under a project, the backbone of Braintrust observability.

Braintrust Prompts API

Create, version, list, update, and delete prompts that bundle model, messages, and parameters for reuse across the platform and at runtime.

Braintrust Functions and Scorers API

Manage and invoke functions - tools, prompts, and code/LLM scorers - including create, list, update, delete, and server-side invocation.

Braintrust Project Configuration API

Configure project-level resources - project scores, tags, automations, saved views, and span iframes - that shape how data is scored and displayed.

Braintrust Organization and ACL API

Manage organizations, members, users, groups, roles, and fine-grained ACLs for governing access to every Braintrust resource.

Braintrust Credentials and Secrets API

List and revoke API keys, manage service tokens, store provider AI secrets, and set environment variables and MCP server connections.

Braintrust AI Proxy API

OpenAI-compatible proxy for chat/completions, completions, and embeddings across many model providers, with caching, logging, and temporary credentials.

Collections

Braintrust REST API

OPEN

Pricing Plans

Braintrust Data Plans Pricing

3 plans

PLANS

Rate Limits

Braintrust Data Rate Limits

7 limits

RATE LIMITS

FinOps

Braintrust Data Finops

FINOPS

Event Specifications

Braintrust Streaming (SSE) API

AsyncAPI description of Braintrust's documented HTTP Server-Sent Events (SSE) streams. Braintrust does NOT expose a WebSocket API. Streaming is delivered as one-way HTTP SSE ove...

ASYNCAPI

Resources

👥

GitHubOrganization

Sources

opencollection: 1.0.0
info:
  name: Braintrust REST API
  version: '1.0'
request:
  auth:
    type: bearer
    token: '{{bearerToken}}'
items:
- info:
    name: Projects
    type: folder
  items:
  - info:
      name: List projects
      type: http
    http:
      method: GET
      url: https://api.braintrust.dev/v1/project
    docs: List out all projects, optionally filtered and paginated.
  - info:
      name: Create project
      type: http
    http:
      method: POST
      url: https://api.braintrust.dev/v1/project
      body:
        type: json
        data: "{\n  \"name\": \"my-project\"\n}"
    docs: Create a new project; returns an existing project of the same name unmodified.
  - info:
      name: Get project
      type: http
    http:
      method: GET
      url: https://api.braintrust.dev/v1/project/{{project_id}}
    docs: Get a project by its unique identifier.
  - info:
      name: Delete project
      type: http
    http:
      method: DELETE
      url: https://api.braintrust.dev/v1/project/{{project_id}}
    docs: Delete a project by its unique identifier.
- info:
    name: Experiments
    type: folder
  items:
  - info:
      name: Create experiment
      type: http
    http:
      method: POST
      url: https://api.braintrust.dev/v1/experiment
      body:
        type: json
        data: "{\n  \"project_id\": \"{{project_id}}\",\n  \"name\": \"my-experiment\"\n}"
    docs: Create a new experiment under a project.
  - info:
      name: Insert experiment events
      type: http
    http:
      method: POST
      url: https://api.braintrust.dev/v1/experiment/{{experiment_id}}/insert
      body:
        type: json
        data: "{\n  \"events\": [\n    {\n      \"input\": {\"question\": \"What is 2+2?\"},\n      \"output\": \"4\",\n \
          \     \"expected\": \"4\",\n      \"scores\": {\"accuracy\": 1}\n    }\n  ]\n}"
    docs: Insert a set of events into the experiment.
  - info:
      name: Fetch experiment events
      type: http
    http:
      method: POST
      url: https://api.braintrust.dev/v1/experiment/{{experiment_id}}/fetch
      body:
        type: json
        data: "{\n  \"limit\": 100\n}"
    docs: Fetch the events in an experiment.
  - info:
      name: Summarize experiment
      type: http
    http:
      method: GET
      url: https://api.braintrust.dev/v1/experiment/{{experiment_id}}/summarize?summarize_scores=true
    docs: Summarize the scores and metrics of an experiment.
- info:
    name: Datasets
    type: folder
  items:
  - info:
      name: Create dataset
      type: http
    http:
      method: POST
      url: https://api.braintrust.dev/v1/dataset
      body:
        type: json
        data: "{\n  \"project_id\": \"{{project_id}}\",\n  \"name\": \"my-dataset\"\n}"
    docs: Create a new dataset under a project.
  - info:
      name: Insert dataset events
      type: http
    http:
      method: POST
      url: https://api.braintrust.dev/v1/dataset/{{dataset_id}}/insert
      body:
        type: json
        data: "{\n  \"events\": [\n    {\n      \"input\": {\"question\": \"What is the capital of France?\"},\n      \"expected\"\
          : \"Paris\"\n    }\n  ]\n}"
    docs: Insert a set of records into the dataset.
  - info:
      name: Fetch dataset events
      type: http
    http:
      method: POST
      url: https://api.braintrust.dev/v1/dataset/{{dataset_id}}/fetch
      body:
        type: json
        data: "{\n  \"limit\": 100\n}"
    docs: Fetch the records in a dataset.
- info:
    name: Logs
    type: folder
  items:
  - info:
      name: Insert project logs
      type: http
    http:
      method: POST
      url: https://api.braintrust.dev/v1/project_logs/{{project_id}}/insert
      body:
        type: json
        data: "{\n  \"events\": [\n    {\n      \"input\": {\"prompt\": \"Hello\"},\n      \"output\": \"Hi there!\",\n  \
          \    \"metadata\": {\"model\": \"gpt-4o\"}\n    }\n  ]\n}"
    docs: Insert trace span / log events into a project's logs.
  - info:
      name: Fetch project logs
      type: http
    http:
      method: POST
      url: https://api.braintrust.dev/v1/project_logs/{{project_id}}/fetch
      body:
        type: json
        data: "{\n  \"limit\": 100\n}"
    docs: Fetch trace span / log events from a project's logs.
- info:
    name: Prompts
    type: folder
  items:
  - info:
      name: List prompts
      type: http
    http:
      method: GET
      url: https://api.braintrust.dev/v1/prompt?project_id={{project_id}}
    docs: List out all prompts, optionally filtered by project or slug.
  - info:
      name: Create prompt
      type: http
    http:
      method: POST
      url: https://api.braintrust.dev/v1/prompt
      body:
        type: json
        data: "{\n  \"project_id\": \"{{project_id}}\",\n  \"name\": \"Greeting\",\n  \"slug\": \"greeting\",\n  \"prompt_data\"\
          : {\"options\": {\"model\": \"gpt-4o\"}}\n}"
    docs: Create a new prompt under a project.
- info:
    name: Functions
    type: folder
  items:
  - info:
      name: List functions
      type: http
    http:
      method: GET
      url: https://api.braintrust.dev/v1/function?project_id={{project_id}}
    docs: List out all functions and scorers, optionally filtered by project.
  - info:
      name: Invoke function
      type: http
    http:
      method: POST
      url: https://api.braintrust.dev/v1/function/{{function_id}}/invoke
      body:
        type: json
        data: "{\n  \"input\": {\"text\": \"Some text to score\"},\n  \"stream\": false\n}"
    docs: Invoke a function (prompt, code, or LLM scorer) server-side. Set stream to true for an SSE stream.
- info:
    name: Evals
    type: folder
  items:
  - info:
      name: Launch eval
      type: http
    http:
      method: POST
      url: https://api.braintrust.dev/v1/eval
      body:
        type: json
        data: "{\n  \"project_id\": \"{{project_id}}\",\n  \"data\": {\"dataset_id\": \"{{dataset_id}}\"},\n  \"task\": \"\
          my-task-function\",\n  \"scores\": [\"Factuality\"]\n}"
    docs: Launch an evaluation that runs a task function over a dataset and applies scorers.
- info:
    name: Organization & ACL
    type: folder
  items:
  - info:
      name: List organizations
      type: http
    http:
      method: GET
      url: https://api.braintrust.dev/v1/organization
    docs: List out all organizations the caller belongs to.
  - info:
      name: List ACLs
      type: http
    http:
      method: GET
      url: https://api.braintrust.dev/v1/acl?object_type=project&object_id={{project_id}}
    docs: List ACLs for a given object type and id.
- info:
    name: Credentials
    type: folder
  items:
  - info:
      name: List API keys
      type: http
    http:
      method: GET
      url: https://api.braintrust.dev/v1/api_key
    docs: List out all API keys (without secret values).
- info:
    name: AI Proxy
    type: folder
  items:
  - info:
      name: Proxy chat/completions
      type: http
    http:
      method: POST
      url: https://api.braintrust.dev/v1/proxy/chat/completions
      body:
        type: json
        data: "{\n  \"model\": \"gpt-4o\",\n  \"messages\": [{\"role\": \"user\", \"content\": \"Hello!\"}],\n  \"stream\"\
          : false\n}"
    docs: OpenAI-compatible chat completions proxied to a model provider. Set stream to true for an SSE stream.
  - info:
      name: Proxy embeddings
      type: http
    http:
      method: POST
      url: https://api.braintrust.dev/v1/proxy/embeddings
      body:
        type: json
        data: "{\n  \"model\": \"text-embedding-3-small\",\n  \"input\": \"The quick brown fox\"\n}"
    docs: OpenAI-compatible embeddings proxied to a model provider.