Braintrust
Evals and production tracing for teams shipping AI agents
Overview
Braintrust is an evaluation and observability platform for LLM applications and agents: trace production traffic, turn failures into datasets, and score prompt or model changes before they ship. Customer stories include Notion, Dropbox and Cloudflare. A free Starter plan has unlimited users; Pro is $249 a month plus metered usage.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 14 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- $10 of model credits a month, then token rates
- 1 GB processed data, then $4/GB
- 10k scores, then $2.50 per 1k
- 14-day retention; unlimited users, projects and experiments
- $100 of model credits a month
- 5 GB processed data, then $3/GB
- 50k scores, then $1.50 per 1k
- 30-day retention, then $0.50/GB/month up to 180 days
- Custom charts, environments, priority support and basic RBAC
- Custom retention and S3 data export
- SAML SSO, custom RBAC, BAA and uptime SLA
- On-prem or hosted deployment; shared Slack channel
Use cases
What you can produce with Braintrust
- Eval suites as code that score LLM outputs against versioned datasets on every prompt or model change
- Side-by-side experiments showing which test cases regressed before a release
- Production traces turned into regression-test dataset rows with one click
- Recurring production failures surfaced by Patterns and topic classification
- Prompts, scorers and datasets drafted by the Loop agent
- LLM-as-a-judge, code and human scoring in one place
- Logs queried and evals run from a coding agent through the MCP server
Braintrust has the most complete trace-to-dataset-to-experiment loop we have seen for teams that treat evaluation as engineering. Loop drafts scorers and datasets, Patterns surfaces recurring production failures, and Brainstore can run in your own infrastructure. The free Starter plan, with unlimited users, is enough to run a real eval workflow.
The catch is cost shape: Pro is $249 a month before metered data and scores, and retention is short below Enterprise. Worth it for product-engineering teams shipping LLM features at volume; overkill for a side project or anyone needing only a trace viewer.
What we found
Braintrust is an evaluation and observability platform for teams putting LLM features and agents into production, run by Braintrust Data, Inc. in San Francisco and led by co-founder and chief executive Ankur Goyal. It launched in 2023. In October 2024 it raised a $36 million Series A led by Andreessen Horowitz, bringing total funding to $45 million, and on 17 February 2026 an $80 million Series B led by ICONIQ. Axios and SiliconANGLE reported an $800 million valuation; Braintrust's own announcement does not state one.
The core loop is unchanged: log production traces, turn a bad one into a dataset row with a click, then run experiments that score prompt or model changes against that dataset with LLM-as-a-judge, code or human scorers. Around it Braintrust has built Loop, an agent that drafts prompts, scorers and datasets; Patterns and Topics, which surface recurring behaviour in production traces; custom trace views for reviewers; an MCP server; and SDKs for Python, TypeScript, Go, Ruby and C#. Traces live in Brainstore, its own database, which can run in a customer's infrastructure as a hybrid deployment. Earlier versions of this page said automatic issue discovery was missing; Patterns now covers it.
Starter is free with unlimited users, $10 of model credits, 1 GB of processed data, 10,000 scores and 14-day retention. Pro is $249 a month with $100 of credits, 5 GB, 50,000 scores and 30-day retention, extendable to 180 days at $0.50 per GB a month. Overages are metered, so cost follows trace volume. Qualifying startups can get Pro free for six to twelve months, and Enterprise adds SAML SSO, a HIPAA business associate agreement and on-premises deployment.
Customer stories on its site include Notion, Dropbox, Cloudflare, Coursera, Box and Retool. A review score and reviewer claims quoted here before could not be traced and have been removed. It suits teams that will actually write scorers; anyone who only wants a trace viewer pays for machinery that will sit unused.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including Braintrust's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Braintrust unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Braintrust directly →
Spotted something out of date? Suggest an update →
More in Code & Dev Tools
