PeerLoom
Local-first engineering · Beta in preparation

Build on your PC.
Stay in control
from your phone.

PeerLoom coordinates Codex and Claude to develop independent solutions, compare their work and validate a final candidate against your project’s configured checks.

Control it from your phone See how it works

Built for Windows. Uses your installed agents and provider accounts.

YOUR PC, FROM YOUR PHONE

Step away.
Stay in control.

Start assignments, follow progress, review changes and approve results from your paired phone, wherever you have internet access, while your Windows PC stays on, awake, signed in and connected.

  • Start and follow work

    Start an assignment, watch Codex and Claude progress, and see the moment your review is needed.

  • Review and decide on the go

    Read the changes, check results, and limitations, then approve the next step or request changes.

  • Paired by you, limited by you

    Pair a phone with a one-time QR invitation, matching codes on both screens, and approval on your PC. Choose what it can do and for which projects, and revoke it any time.

You control what your phone can do. Codex and Claude use your existing provider accounts and send relevant project content to OpenAI and Anthropic for processing. Your project files and agent processes stay on your PC.

FROM INTENT TO EVIDENCE

Your goal starts the work.
The evidence guides the decision.

A conversation gets you started. Structured workflows keep the next steps visible.

01

Bring your project

Import a Git URL, choose a local folder, upload a ZIP on your PC, or start from scratch.

02

Describe one clear goal

Explain the outcome and what must keep working. Review the agents, acceptance checks, and execution limits.

03

Let different approaches emerge

Codex and Claude each build in their own isolated folder. A synthesis step compares both builds and proposes one result.

04

Decide with the evidence

Inspect the changes, saved check results, and limitations. Approve, request changes, or let Auto continue within limits you set.

A SECOND PERSPECTIVE, WITH RECEIPTS

More visibility.
Fewer leaps of faith.

Read the candidate changes and the checks run against the exact result. Agent opinions and test results stay distinct, so you can judge what each one establishes.

WHAT YOU CAN REVIEW
Changes
Exact candidate commits
Checks
Saved validation results
Context
Reasoning and limitations
Control
Approvals and bounded Auto

A completed workflow records a decision. Merging, pushing, and deploying remain separate actions.

WAYS TO THINK IT THROUGH

Choose the depth
your work needs.

Start with Bright. Deeper modes and your own workflows are on the way.

Bright

Beta target

Two builds, one combined result.

  1. Codex and Claude each build your goal on their own, in separate workspaces.
  2. Codex compares both builds, keeps what works best, and produces one final version.
  3. Your checks run against that exact result. You review the evidence and decide.

Beyond Bright

Brighter

Internal preview

An extra refinement round before the final version.

Brilliant

Internal preview

An independent review of the finished changes.

Your own workflow

Preview

Arrange the agents, roles, and review steps to fit your project.

Modes don’t guarantee a particular cost or quality. Every assignment has call and time limits you review before it starts; they are not price estimates.

WHERE THINGS STAND

Private beta
in preparation.

The first beta focuses on Windows, with Codex and Claude in Bright mode. Public downloads and open signup are not available yet. Merging, pushing and deploying remain separate actions.

Beta testers use their own Codex and Claude subscriptions, so agent usage counts against their own plans.

FIRST BETA SCOPE
Platform
Windows
Agents
Your installed Codex and Claude
Workflow
Bright
Workspace
Runs on your PC
Phone
Paired access from anywhere
Access
By invitation