Braintrust - AI Tools Nav

What it is for

Braintrust is an eval + trace workbench: you log a run, score it (code, LLM-as-judge, or a human), and compare experiments in one project. Use it when the job is measure the agent before you ship, not when you only need a request log.

It is not open source as a product. Starter is $0/month (usage overages apply). Pro is $249/month.

What makes it unique: the cover is Braintrust’s official plan grid — Starter $0/month, Pro $249/month, credits / GB / scores on the same cards you checkout.

Key features

Tracing — log spans from your app and inspect them next to the eval that used the same data (Plans).
Experiments — compare prompt / model / scorer changes as rows you can ship or reject (Pricing).
Scores — LLM-as-judge, autoevals, or custom code; Starter includes 10k scores / month, then $2.50/1k (Pricing).
Playgrounds — iterate a prompt against a dataset without redeploying (Pricing).
Loop agent — Braintrust’s built-in agent that can run evals and propose test cases (Pricing).
Human review — configurable annotation scores your team fills in on a trace (Pricing).

Who it's for

Best fit: AI teams who will write a scorer and pay when processed data or scores blow past Starter.

Poor fit: developers who only want a $0 request table (Helicone).

Pricing

Verified 2026-10-22 (Asia/Shanghai) from braintrust.dev/pricing and Plans and limits. Starter launched as the replacement Free path (March 16, 2026 blog). Re-check Billing.

Starter — $0/month. $10 model credits + token rates. 1 GB processed data, then $4/GB. 10k scores, then $2.50/1k. 14-day retention. Unlimited users, projects, datasets, playgrounds, experiments. No credit card.
Pro — $249/month. $100 credits. 5 GB data, then $3/GB. 50k scores, then $1.50/1k. 30-day retention, then $0.50/GB/month. Custom charts, environments, RBAC, priority support.
Enterprise — custom. On-prem / hosted, custom retention / export, compliance.

Model credits do not roll. Built-in models after credits bill at the printed token rates (Topics $0.06 / $0.40 per MTok on that table). Early-stage startups can apply for 6–12 months of Pro.

Limitations / costs

$0 is the platform fee. Data and scores still meter.
14-day Starter retention. A weekly eval habit wants Pro or an export.
Credits reset every month.
This is not a gateway. You still pay the model vendor (or Braintrust’s built-in rates).

Alternatives

Langfuse — open-source traces + scores if you will self-host.
LangSmith — LangChain-native experiments.
LangWatch — agent simulations + traces with a €29 Growth seat.

Official tutorial

Official Braintrust channel video: How to evaluate AI agents with Braintrust (oEmbed author: Braintrust).

Written path: Pricing plus Plans and limits.

Website

https://www.braintrust.dev

Source

Catalog row curated for AI Tools Nav. Cover is Braintrust’s official Starter $0 / Pro $249 plan cards. Tutorial: evaluate agents on @BraintrustData (oEmbed).

Last verified

2026-10-22 (Asia/Shanghai) editorial check.