Introduction
Sync is a desktop app and knowledge layer for AI coding agents. Your agent reads and writes structured docs in your repo — and every document tells you, and the agent, the moment it goes stale.
Sync is a desktop app for working with AI coding agents — backed by a knowledge layer that lives in your repo. You open a repository, run your agent inside Sync, and it reads the project's specs, decisions, and constraints before it writes code, and records what it learns as it works.
The one idea that makes it different: every document carries a status. When the code changes, Sync marks the documents that describe that code as stale — so your agent always knows whether what it's reading is still true.
What you actually do
- Open your repo in the app and run your agent in the built-in terminal. Click a spec, say "implement this," and the agent already knows what you mean.
- Browse knowledge and the roadmap visually — no files to hunt through.
- Answer the agent's questions and review what it writes inline, like comments on a pull request.
- Keep steering. You set the plan and the priorities; the agent works against them, and you stay in control of where a big task is heading.
Start here
Quickstart
Download the app, open a repo, and run your first session.
How staleness works
The core idea: how a document's status tells your agent it has gone out of date.
What Sync tracks
Specs, decisions, constraints, and the roadmap — at a glance.
Connect your agent
Use Sync with Claude Code, Cursor, Zed, Codex, or OpenCode.
Good to know
- Local-first. Everything lives in your repo, in git. Nothing leaves your
machine unless you
git pushit. No telemetry. - Open source. Released under FSL-1.1-MIT, converting to MIT after two years. Source: github.com/sync-buzz/sync.
- Alpha. Still stabilizing — expect some rough edges and the occasional format migration before 1.0.