From the experiment

Do coding agents recommend Braintrust?

Braintrust was chosen in 6% of 288 judged evals sessions, ranking fourth. Measured with Claude Code, Codex and Cursor.

Published September 3, 2026 Read as Markdown

Braintrust was chosen in 6% of 288 judged evals sessions, ranking fourth. It was also raised as a candidate in 207 further sessions without being chosen.

This page reports what happened when Claude Code, Codex and Cursor had to solve a problem in evals inside a realistic codebase. Not what a chat assistant says about Braintrust. What an agent actually installed.

The numbers

CategoryEvals
Sessions in the category288
Sessions where Braintrust was chosen18
Install share6%
Rank in category4 of 9
Codebases it won in2
Raised as a candidate, not chosen207
Chosen when considered8%
Sitebraintrust.dev

By agent

With 18 wins spread across three agents, the rates below are small numbers and a difference between them is not yet a finding. They are here because the direction is worth knowing, not because the gap is established.

AgentSessionsChose BraintrustShare
Claude Code9600%
Codex961718%
Cursor9611%

By who was asking

Braintrust performs similarly across the four kinds of buyer, from 2% to 10%. That is unusual: the category leader changed with the persona in 14 of the 18 categories we measured.

Who is askingSessionsChose BraintrustShare
Junior developer14532%
Senior engineer1431510%

What Braintrust was up against

The full ranking in evals, from the same sessions:

#ProductRuns wonShare
1Langfuse9734%
2Built in-house (no product adopted)8429%
3Promptfoo3813%
4Arize Phoenix238%
5Braintrust (this page)186%
6Inspect AI124%
7LangSmith83%
8Helicone62%

What this means

This is an integration problem, not a presence problem. Agents raised Braintrust in 207 sessions and chose it in 18, so it reaches the shortlist and then loses. Something at the last step is costing the session, and in our data that is usually a quickstart that does not run when pasted, documentation describing an interface that changed, or a package name that does not match the product name.

That is the cheaper of the two problems to have. The reason is written down in each losing transcript.

Where these numbers come from

The 288 sessions in evals are part of a published set of 5,292, run with real coding agents inside realistic codebases and judged blind. The full method is on one page: how we measured this.

Every evals run can be replayed on the board.

If you work on Braintrust: the judge recorded a reason for every session where it was raised and passed over. Those reasons are in the transcripts.

<!-- generated by scripts/write-data-pages.mjs -->

Common questions

Do coding agents recommend Braintrust?

Yes. Braintrust was chosen in 18 of the 288 judged sessions in evals, a 6% install share, ranking fourth in its category.

Does Claude Code recommend Braintrust?

In 0 of the 96 sessions in evals run with Claude Code, which is 0%.

Do different coding agents treat Braintrust differently?

Yes, and by a wide margin. Codex chose it in 18% of its runs and Claude Code in 0%.

How was this measured?

Real coding agents at pinned versions were run in sandboxes inside 51 realistic codebases and asked to solve real tasks. A simulated project owner approved or questioned each recommendation before any code was written, and a judge from a model family that builds none of the agents read every session blind.

How often is Braintrust considered but not chosen?

It was raised as a candidate in 207 sessions without being chosen, and chosen in 18. That is a 8% conversion from considered to chosen.

Where this comes from

Armature ran 5,292 judged sessions with Claude Code, Codex and Cursor inside 51 realistic codebases, and published every run. The numbers on this page come from that work.

Read next

All library pages