Braintrust
Braintrust is an AI-native observability and evaluation platform that helps engineering, product, and data teams at enterprises measure, debug, and improve AI products in production through tracing, scoring, and automated pattern discovery, with named customers including Notion, Stripe, Vercel, Dropbox, Replit, and Cloudflare.
- Company typePrivate
- Founded2023
- HeadquartersSan Francisco, United States
- Headcount101–250
- GTM typeB2B
- OfferingSoftware
What Braintrust does
Braintrust Data, Inc. is an AI-native observability and evaluation platform founded in 2023 by Ankur Goyal, the former head of AI at Figma and founder of Impira. Headquartered in San Francisco, the company provides a single platform that connects production tracing, output quality evaluation, and systematic iteration for AI products, framing the category as "eval-driven development" infrastructure for engineering, product, and data teams. Its technology stack centers on Brainstore, a proprietary AI-native database optimized for large, nested, semi-structured AI trace data, and is delivered through native SDKs in Python, TypeScript, Go, Java, Ruby, and C# with auto-instrumentation across more than 40 LLM providers, agent frameworks, and coding agents. Named modules include Loop (a generative-AI assistant that writes prompts, scorers, and datasets using Claude 4), Topics (ML-driven log clustering for production pattern discovery), the Braintrust Gateway (a unified multi-provider LLM API with caching, observability, and workload identity federation), the Braintrust CLI, and a Model Context Protocol server for IDE-based coding agents.
Braintrust serves a blue-chip enterprise customer base including Vercel, Notion, Stripe, Dropbox, Replit, Cloudflare, Ramp, Airtable, Instacart, Coursera, Coda, Zapier, Navan, and Graphite, with cited use cases in RAG observability, agent monitoring, prompt experimentation, hallucination detection, and CI/CD quality gates. Its go-to-market is hybrid: a free self-serve tier with a no-code Playground for product managers is paired with a dedicated enterprise field-sales organization covering US (San Francisco, New York, Chicago) and EMEA (London) regions. Pricing combines monthly SaaS subscriptions (Starter, Pro, Enterprise) with usage-based overage on AI provider tokens and Topics processing, and the company holds SOC 2 Type II, HIPAA, and GDPR certifications with SSO/SAML, granular RBAC, and hybrid data-plane deployment for regulated buyers.
Braintrust has raised approximately $121M in disclosed equity funding across three rounds, culminating in an $80M Series B in February 2026 led by ICONIQ Capital at an $800M post-money valuation, with continued participation from Andreessen Horowitz, Greylock, Elad Gil, and Basecase Capital. ICONIQ General Partner Matt Jacobson joined the board as part of the transaction. New Market Pitch ranks Braintrust as the #1 AI governance startup by valuation and #2 by total funds raised as of 2026. In May 2026 the company disclosed unauthorized access to an AWS account holding customer API keys, prompting a credential-rotation request to all customers.
Braintrust firmographics
Firmographics- Name
- Braintrust
- Legal name
- Braintrust Data, Inc.
- Website
- https://braintrust.dev
- Company type
- Private
- Founded year
- 2023
- Operating status
- Operating
- Headcount range
- 101–250 employees
- Short description
- Braintrust is an AI-native observability and evaluation platform that helps engineering, product, and data teams at enterprises measure, debug, and improve AI products in production through tracing, scoring, and automated pattern discovery, with named customers including Notion, Stripe, Vercel, Dropbox, Replit, and Cloudflare.
- Ownership category
- akta.pro rank
Where Braintrust is headquartered
LocationHeadquarters
- HQ city
- San Francisco
- HQ country
- United States
- HQ region
- North America
Offices3 records
Markets served
Braintrust business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Personnel, Technology or R&D, Marketing or Sales, Infrastructure, Operations
Revenue model
- Subscription (SaaS plans): Recurring SaaS subscriptions across Starter, Pro, and Enterprise plans. Topics is priced via monthly included credits ($10/month on Starter, $249/month on Pro during launch promotion for first 3 months, reverting to $100/month). Pro and Enterprise tiers required for premium features (dataset snapshots, sandboxes for agent evals, annotating playground outputs).
- Usage-based overage charges: Usage-based component for AI provider tokens processed via the Braintrust Gateway and Topics. Overage charges apply once included monthly credits are exhausted, at $0.06/MTok input and $0.40/MTok output, uniform across plans. Credits do not roll over month-to-month.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Freemium | Monthly | Free entry tier plus Starter, Pro, and Enterprise plans |
| Usage-based | Pay-as-you-go | Usage-based overage on AI provider tokens and Topics processing |
Go-to-market motion4 records
Distribution channels5 records
Marketing channels10 records
Braintrust product offering
Product offeringCore offering
Braintrust is a unified AI observability and evaluation platform that connects production tracing, output-quality evaluation, and systematic experimentation in a single workflow. It helps AI engineering, product, and data science teams inspect prompts, responses, and tool calls in real time, score outputs with LLMs/code/humans against golden datasets, and automate pattern discovery, online scoring, and CI/CD quality gates to ship quality AI at scale.
Product overview
Braintrust is a single unified AI observability and evaluation platform rather than separate point products: it connects tracing, evaluation, and experimentation in one workflow through three pillars (Observability, Evals, Automation) and a shared data plane. The core platform is augmented by named modules — Brainstore (the AI-native database for traces), Loop (generative-AI assistant that writes prompts, scorers, and datasets), Topics (automated log-insights clustering), the Braintrust Gateway (unified multi-provider LLM API), the Braintrust CLI (bt) for terminal workflows, the Braintrust MCP Server for IDE-based coding agents, and the Playground no-code interface — all of which share Brainstore, the eval framework, and SDK instrumentation libraries (Python, TypeScript, Go, Java, Ruby, C#, Autoevals).
Differentiator
Problem solved
Functional benefit
Brands
- Brainstore: Database built for AI data at scale, designed for complex AI traces and AI observability workloads.
- Loop
Products and services
- Braintrust Platform The unified AI observability and evaluation platform that connects tracing, evaluation, and experimentation in a single workflow. Provides observability (real-time tracing of prompts, responses, and tool calls), evals (scoring outputs with LLMs, code, or humans against golden datasets), and automation (Topics, online scoring, quality gates) to ship quality AI at scale. For engineering, product, and data science teams building AI products in production.
- Brainstore Proprietary database built specifically for AI observability, designed to handle large, nested AI traces that traditional databases cannot efficiently process. Optimized for full-text search, write latency, and span load acceleration with hybrid deployment on customer infrastructure. For engineering teams running AI in production.
- Loop AI assistant that helps users improve AI: generates better prompts, scorers, and datasets from natural-language descriptions, automates prompt optimization and dataset generation, and is available on project overviews and individual trace pages for natural-language analysis. For AI engineers and product teams using Braintrust.
- Topics Automated log-insights engine that classifies production traces by task, sentiment, and issues using ML clustering. Supports custom topic maps, custom facets, and session-level grouping, and feeds classifications into eval datasets, scorers, review queues, and online scoring rules. For AI engineering teams monitoring production AI behavior.
- Braintrust Gateway Production-grade unified LLM API that routes requests across OpenAI, Anthropic, Google, AWS, Mistral, and third-party providers, with automatic caching, observability, billing, and workload-identity-federated authentication. Supports OpenAI-compatible and Anthropic-compatible SDKs interchangeably, plus native Gemini embedding endpoints. For engineering teams building production LLM applications.
- Braintrust CLI (bt) Command-line interface for terminal access to Braintrust features: running evaluations, querying logs with SQL, managing prompts and projects, syncing data to local NDJSON files, configuring coding agents, push/pull of datasets/functions, and snapshot management. For developers and AI engineers using Braintrust.
- Braintrust MCP Server Model Context Protocol server that lets coding agents (Cursor, Claude Code, OpenCode, pi) query logs, run evals, and update prompts directly from their IDE. Includes resources for auto-installing the Braintrust SDK and a sql_query tool for log SQL queries. For AI engineers using coding agents.
- Playground Visual no-code UI within Braintrust that lets product managers and engineers experiment with prompts, models, and settings, run evaluations, annotate outputs, share experiment links, and turn reviewed examples into golden datasets without writing code. For product managers and engineers iterating on AI features.
Quantifiable outcome
- 45x more feedback with AI grading at Coursera
- +4 more outcomes
Companies that use Braintrust
Customer profileNamed customers14 records
Segments5 records
Ideal customer profiles4 records
Braintrust technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
Integration47 records
AI capability15 records
Feature7 records
Braintrust partnerships and signals
Strategic signalPartnerships
Twelve partnerships are on record, tiered flagship and strategic.
- AnthropicflagshipDeep technology integration: Braintrust supports workload identity federation for Anthropic (May 2026) and AI Gateway for cross-SDK routing (call Claude through OpenAI SDK). Anthropic instrumentation captures per-TTL prompt cache metrics, time_to_first_token, beta messages tools, and extended thinking. Braintrust Loop leverages Claude 4 for automated eval optimization.
- Google (Gemini / Vertex AI)flagshipDeep technology integration with Google Gemini and Vertex AI for tracing, embeddings (text + multimodal image), workload identity federation (April/May 2026), and AI Gateway routing. Supports Gemini thinking/reasoning, Live API, images, generateContent, streamGenerateContent, embedContent, and batchEmbedContents.
- OpenAIflagshipDeep technology integration: Braintrust supports workload identity federation for OpenAI (exchanging Braintrust-signed JWTs for OpenAI access tokens) and offers an AI Gateway that lets users call OpenAI models through the Anthropic SDK or Google SDK. OpenAI integration supports Responses API, streaming, audio transcription, embeddings, image generation, and OpenAI Agents SDK tracing.
- Mistral AIstrategicTechnology integration with Mistral for native tracing of chat, streaming, embed, audio speech/transcription, OCR, moderation, classification, and beta.conversations tool calls. Mistral models available through Braintrust AI Gateway.
- Hugging FacestrategicTechnology integration with Hugging Face Hub for tracing InferenceClient calls (chat completion, text generation, feature extraction, and sentence similarity).
- OpenRouterstrategicTechnology integration with OpenRouter for native SDK tracing, AI Gateway routing, and OpenRouter Agent SDK tracing. Supports reasoning content capture.
- Anthropic (Claude Agent SDK)strategicDeep integration with Claude Agent SDK for tracing hooks, CLI usage, parallel child spans, sub-agent nesting, and lifecycle handling. Supports embedding Claude Code traces inside parent traces.
- CoherestrategicNative tracing integration with Cohere SDK covering v1/v2 chat, streaming, embed, rerank, audio transcription, classification, and moderation APIs. Captures extended thinking and reasoning_tokens.
- Microsoft Azure (Azure OpenAI / Azure AI Foundry)strategicTechnology integration with Azure OpenAI and Azure AI Foundry via Workload Identity Federation using Microsoft Entra Workload Identity (June 2026). GPT-5.5 Pro available through Azure via Braintrust gateway.
- VercelstrategicVercel AI SDK is a deeply integrated framework partner with auto-instrumentation for Vercel AI SDK v3-v6. Braintrust provides braintrustVitePlugin, braintrustWebpackPlugin, braintrustEsbuildPlugin, braintrustRollupPlugin, and wrapNextjsConfigWithBraintrust for Next.js Turbopack support. Vercel is also a customer (CTO Malte Ubl publicly endorses Braintrust).
- Amazon Web Services (Bedrock)flagshipDeep technology integration with AWS Bedrock for tracing (converse and converseStream), assume role authentication via AWS STS, and hybrid deployment options. Bedrock custom models can route through Bedrock's OpenAI-compatible Chat Completions endpoint. Bedrock custom models are also supported.
- LangChain / LangGraphstrategicNative integration with LangChain (Python and TypeScript) and LangGraph for auto-instrumentation. Supports embedding tracing. Braintrust complements LangChain ecosystem with observability and eval tooling.
Scale indicators9 records
Recent moves6 records
Expansion highlights7 records
Braintrust competitors and assessment
Company assessmentDirect peers
- LangSmith: LangChain's commercial observability and evaluation platform for LLM applications. Directly competes with Braintrust across tracing, evals, dataset management, and prompt iteration, with a distribution advantage from the dominant LangChain framework.
- Arize AI: AI observability and LLM evaluation platform with the open-source Phoenix project. Competes head-to-head with Braintrust on tracing, evals, and drift monitoring for production AI; ranked ahead of Braintrust by total fundraising ($131M vs $121M) per New Market Pitch 2026.
- Langfuse: Open-source LLM engineering platform for tracing, evals, prompt management, and datasets. Competes with Braintrust on the same core workflows with an OSS-first GTM motion that pressures pricing.
- Galileo: Evaluation and observability platform purpose-built for generative AI applications. Competes with Braintrust on hallucination detection, RAG evaluation, and agent-quality scoring for enterprise AI teams.
- WhyLabs: AI observability and monitoring platform that provides data and model quality monitoring, including LLM-specific capabilities. Competes with Braintrust on production monitoring for AI applications.
- Helicone: Open-source LLM observability platform offering logging, monitoring, and an AI gateway with caching and routing. Overlaps with Braintrust's Gateway and tracing offerings at a developer-first price point.
Emerging players
- Maxim AI: Evaluation and observability platform for AI agents and LLM applications. Competes with Braintrust on agent evaluation, simulation, and production tracing, with overlap in eval-driven development workflows.
- Confident AI (DeepEval): Open-source LLM evaluation framework (DeepEval) with a hosted platform for AI testing and observability. Competes with Braintrust on evals, particularly for teams that prefer open-source-first tooling.
Broad incumbents
- Datadog: Established cloud observability platform that has added LLM observability capabilities and participated in Braintrust's Series A. Represents the incumbent threat of bundling AI evaluation into a broader monitoring stack.
- Honeycomb: Observability platform for production engineering with event-based tracing and high-cardinality analytics. A broader observability peer that Braintrust references as adjacent, and that some customers could pair with or substitute for AI-specific tooling.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat6 records
Key risks5 records
Key highlights7 records
Customer concentration
Braintrust social profiles
Digital presenceBraintrust compliance and trust
Trust signalCompliance3 records
Braintrust financial estimates
Financial estimateRevenue estimate
Valuation estimate
Braintrust leadership team
Management profileNumber of profiles
Profiles3 records
Braintrust funding detail
Funding detailFunding overview
Funding rounds3 records
Investors8 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Braintrust M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about Braintrust
What does Braintrust do?
Braintrust is a unified AI observability and evaluation platform that connects production tracing, output-quality evaluation, and systematic experimentation in a single workflow. It helps AI engineering, product, and data science teams inspect prompts, responses, and tool calls in real time, score outputs with LLMs/code/humans against golden datasets, and automate pattern discovery, online scoring, and CI/CD quality gates to ship quality AI at scale.
Is Braintrust a public or private company?
Braintrust is a private company. It is classified as venture growth investor backed and is currently operating.
When was Braintrust founded?
Braintrust was founded in 2023. It employs 101 to 250 people.
Where is Braintrust based?
Braintrust is headquartered in San Francisco, United States, in the North America region.
How does Braintrust make money?
Two revenue lines are on record. Subscription (SaaS plans) is the primary driver. The others are usage-based overage charges.
Who are Braintrust's main competitors?
Direct peers on record are LangSmith, Arize AI, Langfuse, Galileo, WhyLabs and Helicone. Emerging players are Maxim AI and Confident AI (DeepEval). Broad incumbents are Datadog and Honeycomb.
Does Braintrust have an API?
Yes. Braintrust exposes a public REST API (e.g., POST /v1/service_token, PATCH /v1/organization/members, POST /v1/prompt, PUT /v1/prompt, GET /v1/env_var, GET /v1/ai_secret) for programmatic management of AI providers, prompts, environments, datasets, experiments, and service tokens. Authentication is via Braintrust API keys (organization-level service tokens) and the API also supports webhooks for alerts and automations. SDKs are available in Python, TypeScript/JavaScript, Go, Java, Ruby, and C# for client-side integration. A public Braintrust MCP server is available for connecting coding agents (Cursor, Claude Code, OpenCode, pi) to query logs, run evals, and update prompts directly from an IDE. Developer documentation is at www.braintrust.dev/docs/reference.