LetThemBuild · for Mac, Windows and Linux

Four AIs build it.
Three of them check it.

One leads. Three review the work, run checks, and show their evidence. All in your own checkout — Claude, ChatGPT, Gemini and Grok out of the box, or any of hundreds more through your own endpoint.

Free plan · No card required · $2 of credit to start

LetThemBuildExample run

Add rate limiting

NAYouSep 22, 6:57 AM

Add rate limiting to the upload endpoint and prove it holds.

  • 1ClaudeDraftAdds limit middleware with tests.6:58 AM
  • 2ChatGPTAgreesRan the unit and integration tests.6:59 AM
  • 3GeminiObjectionA burst of requests still gets through.7:00 AM
  • 4ClaudeRevisedCounts per window, not per second.7:02 AM
  • 5GrokAgreesRe-ran the tests. All checks pass.7:04 AM
3 reviewers agree · 7 checks passedView full receipt →

Your models. One workspace.

Bring together the best AIs, or use your own endpoint.

  • Claude
  • ChatGPT
  • Gemini
  • Grok
  • Your own endpoint

01 / From request to reviewed change

Watch the work. See the proof.

  1. 01

    Ask

    Tell the lead what to build, with files or pasted context when it needs them.

  2. 02

    Build

    The lead AI writes the code in your workspace, with real tools and commands.

  3. 03

    Review

    The other AIs read the changes, run checks, and give their verdicts.

  4. 04

    Resolve

    They converge — or call out exactly what to fix. You get the full discussion.

  5. 05

    Your PR

    When you’re happy, merge or discard. The branch is yours.

Every review
comes with a receipt.

The commands each reviewer ran, and their exit codes, on the receipt under every answer. You see exactly what they checked, what passed, and what didn’t. “Looks good to me” doesn’t count: a reviewer that ran nothing is marked as one that ran nothing.

Learn more in the docs

Illustrative review receipt

ReviewerCommandOutcomeDetails
Claude· leadnpm run buildPassed0 errors, 12s
ChatGPT· reviewnpm testPassed42 tests, 8s
Gemini· reviewnpm run lintPassed0 issues, 3s
Grok· reviewnpm run typecheckPassed0 errors, 6s

02 / Built around your workflow

Works the way you do.

Your code. Your machine.

The group works in a git worktree on macOS, Windows or Linux. Nothing leaves your machine unless you choose Our Keys, and the app says so before you do.

▾ MyProject

src

tests

package.json

$ ltb run "add rate limiting"

› lead: building…

› reviewers: running checks…

› agreed ✓ exit 0

Your keys. Your CLI.

Use our keys, bring your own, run on the Claude Code, Codex, Gemini CLI and Grok Build subscriptions you already pay for, or connect any OpenAI-compatible endpoint.

  • Our KeysDefault
  • Your CLI
  • BYOK
  • OpenAI-compatible endpoint

Choose how hard they work.

Quick for a question, Consensus for a change that matters, Deep when it has to be right. Or leave it on Default and the room decides from what you asked and what the lead actually did.

QuickCompareConsensusDeep
  • Quick · one model, one pass
  • Compare · two models, one checking the other
  • Consensus · everyone, until they agree
  • Deep · the same room, twice as long

Build at your desk.
Steer from anywhere.

LetThemBuild is the desktop app where the group builds. LetThemChat, on the web and iPhone, is the same account and the same sessions — for reading what they did, asking the group a question, and steering the next step.

LetThemBuild

Improve error handling

6:53 AM
  • 1ClaudeRevised
  • 2ChatGPTAgrees
  • 3GeminiObjection
  • 4GrokAgrees

‹ LetThemChat

Improve error handling

  • ClaudeRevised
  • ChatGPTAgrees
  • GeminiObjection
  • GrokAgrees
Ask a follow-up…➤

Read how it actually works

The documentation covers the whole product: how the room reaches agreement, what a reviewer has to do before it may agree, what the app is allowed to touch on your machine, and how the money works. No sign-up to read it.