rb. Rabbit Brain

Know where your research stands.

Your AI agent records what was tried, what the results show, and what's still open, with the code and settings.

  1. 1 Open Claude Code or Codex in your project
  2. 2 Paste the prompt
  3. 3 Run rb open to see the record
read the promptshow the prompt to copy
Use Rabbit Brain to keep track of the research in this repository. Record where the work stands now, from its own files.

Set up:
- If `rb --version` fails, install it with `uv tool install rabbit-brain` (or `pipx install rabbit-brain`).
- If this repository has no `.rb/` yet, run `rb init "<the work's theme>"`, then add the question the work is trying to answer with `rb question add`.
- Run `rb docs --start` and follow it. Run `rb context --model <your model id>` before anything else.

Record from the repository's own files, and cite them: the questions, the experiments with each setting and where its value came from, the rules a result must meet, and the runs. Run new experiments through `rb journal run`. Never compute findings, rankings or regressions yourself: rb computes them. Calls on results are mine. Record one only when I ask, in my words, and give me the `rb confirm` command.

If I want a page to share, run `rb page`: it writes the record's page as one file. If I want it in my rabbitbrain.ai account, run `rb login` (I approve its code myself) and then `rb sync`.

Finish with `rb status` exactly as printed, what you recorded, what rb says needs me, and where the page is if you wrote one.

How it works

Paste one prompt. Your agent installs rb and starts the record.

Your agent's chatin your repo

Use Rabbit Brain to keep track of this research: is INT8 ResNet-50 as good as FP32 on ImageNet?
+ 11 more lines
~/int8-resnet

What your agent runsfor that question

  1. rb init "INT8 ResNet-50"creates the .rb/ folder that holds the record
  2. rb question add "Is INT8 ResNet-50 as good as FP32 on ImageNet?"writes down what you want to find out
  3. rb criterion add "INT8 adds at most 0.05 validation loss" ...sets the pass mark before any results exist
  4. rb run add int8 --from out/seed1/metrics.json --set seed=1records each run from its metrics file
  5. rb statusshows the results so far and what waits for your decision

Every result points to the runs and files behind it.

loss-budgetrule ≤ 0.05

0 0.05

criterion met · loss
change.val_loss: +0.01337 [+0.006287, +0.02045]
paired difference, t interval, 3 seeds, 95% (rb)

Your pass mark, checked on every seed

Loss may rise by at most 0.05. Across 3 seeds it rose 0.013, so it passes.

Receipts · rb tour3 runs

  • rundnvs2dout/s1/metrics.jsonsha256 d9be87e8fd4a
  • runkusq34out/s2/metrics.jsonsha256 6cb0e5ba7a76
  • runzdrdumout/s3/metrics.jsonsha256 7691b6ed8f71

config out/s1/config.yaml: 2 match, 0 differ, 0 absent

Each result keeps the files it came from

rb stores a fingerprint of each run's metrics file, so a later change shows. It checks out/s1/config.yaml against your settings: 2 match, 0 differ.

Your callloss-budget

criterion met · loss
acceptrejectinvestigate

$ rb decide loss-budget accept \ --why "within budget on all three seeds" dp3kmue: accepted loss-budget: within budget on all three seeds

You accept or reject a result, with your reason

If your agent accepts or rejects one, it waits until you confirm.

A linkrabbitbrain.ai/demo

rabbitbrain.ai/demo/think-step-by-step-a-small-model-on-grade-school-math
Sample · read-onlysynced 28 Oct 07:57 UTC

A real study, run by an agent and recorded with rb.

Share the whole study with one link

The first three cards show output from rb tour, a small example project rb builds for you.

Under the hood

How rb works, step by step.

  1. [1]
    Your agent writes to the record as it works.It runs experiments with rb journal run, sets pass marks with rb criterion add and leaves notes with rb note add.
  2. [2]
    rb checks the record against your repository.Each setting against your config files, each result against its output file, each run against its git commit. It never edits your code.
  3. [3]
    The next session starts from the record, not from old chats.rb context hands a new agent the questions, results and open problems. You read the same record in rb open, rb page or a shared link.
  4. [4]
    You make the final decisions.If your agent accepts or rejects a result, it stays pending until you run rb confirm.

Checked, not taken on trust

The record doesn't take the agent's word.

  • Settings are checked against your files

    When the agent records a setting, rb looks it up in the config file it cites. If it isn't there, rb marks the setting as not checked.

  • rb redoes the math from the runs

    The sample study cites the model's published score, 49.6%. Its own runs score the plain prompt at 13.1%, so rb marks that number as not reproduced.

  • Late pass marks are flagged

    If a pass mark was written after the scores came in, rb says so next to the result.

  • Every decision keeps who made it and why

    Accept, reject, park or ship: each one stores a name and a reason.

Share

Share a study with one link.

rb sync puts the record in your private account on rabbitbrain.ai. You decide who gets the link, and you can turn it off.

  1. .rb/ in your repo
  2. rb sync --with-record
  3. a link you can revoke
  4. Rerun this research

Free and open source · plain files in your repo

Quick start: a sample study

Try Rabbit Brain on a sample study, so you don't need to set up any experiments or use your own research. Just open Claude Code or Codex in an empty folder and paste the prompt below.

Clone https://github.com/rabbit-brain/think-step-by-step and work inside it.
Install rabbit-brain from PyPI into a virtual environment in that folder.
It's open source: https://github.com/rabbit-brain/rb
Start rb login in the background and show me the sign-in link it prints. I'll sign in.
When I'm signed in, run rb sync and give me the link it prints.
Run rb context and tell me in three lines what the study found and what's still open.
Don't run any experiments.

  1. [1]
    Copy the prompt into Claude Code or Codex.Your agent clones the sample study and installs rb in a virtual environment.
  2. [2]
    Open the sign-in link your agent gives you.An email code, no password.
  3. [3]
    Your own copy of the study opens on rabbitbrain.ai.Private to your account. No account? Ask your agent to run rb page instead.

To use rb on your own repository, start with the install guide.