Works with Claude Code

One file. One agent.
Zero conflicts.

The coordination layer for AI coding agents. Predictive file locking that prevents merge conflicts before they happen.

Get Started→View on GitHub
$npm i -g preclaim

How it works

Three steps to
coordination

01

Install & init

npm i -g preclaimpreclaim init

One command sets up auth, creates your project, and installs Claude Code hooks.

02

Code as usual

Your AI agent edits files normally. Preclaim intercepts writes and atomically locks each file — invisible until it matters.

03

Collaborate safely

Multiple agents, one codebase. Conflicts are prevented at the source — not resolved after the fact.

Features

Built for the
AI-native workflow

claim_file()

Atomic Locking

Database-level file locking via Supabase RPC. One file, one agent, guaranteed.

failOpen: true

Fail-Open

Network down? Development continues. Preclaim never blocks your workflow.

git commit →

Auto-Release

Commit triggers unlock. Crashed sessions expire gracefully via TTL.

subscribe()

Real-time

See active locks across all sessions. Live updates via Supabase Realtime.

PreToolUse

Claude Code Native

First-class hooks integration. Installs in one command, works invisibly.

preclaim init

Self-Service

One command creates your org, project, and auth. No forms, no waiting.

In action

Conflicts stopped
before they start

preclaim — zsh
$ preclaim status
project acme-app
session claude_a8f3c2
ACTIVE LOCKS
────────────────────────────────────────
✓ src/api/auth.ts you 28m
✓ src/lib/database.ts you 25m
✗ src/hooks/useUser.ts claude_b conflict
$ █

Start coordinating
in 30 seconds

$ npm i -g preclaim && preclaim init
View Documentation→