Humans and agents · Working as peers

Your Project Is a Place
Your AI Can Actually Live

Every SyncEngine project is a real folder on your own device. Point a coding agent at it — Claude Code, Codex, or a local model — and it doesn't get a cramped chat window; it gets the whole project. Because to SyncEngine, an agent is a peer.

A Peer, Not a Chatbot

The Folder Is
the Interface

A SyncEngine project isn't a document trapped in a cloud app — it's a real, materialized folder on your machine. That means an AI agent can live inside it: read the whole vault for context, edit files directly, and have its changes flow to your peers exactly the way theirs flow to you. Not a chat sidebar bolted onto a product — a peer in the project.

No uploading your code to someone's servers to "let the AI see it." No copy-pasting between a chatbot and your editor. The agent works where your work already lives — encrypted, on your machine, with no company in the middle.

01 · Attach
Add an Agent

Add an agent to a project. SyncEngine forks the project into its own working folder — the agent's private copy of everything.

02 · Task
Tell It What You Want

Your coding agent runs in that folder with the full project as context — every document, every artifact, the whole vault.

03 · Watch
See It Working

The project shows the agent thinking, then flags when it has changes — you see its work land in real time, like a teammate typing in a shared doc.

04 · Land
Accept With a Click

It merges into the project and syncs to your peers. Nothing half-finished ever leaks — the agent's messy middle stays invisible until it's ready.

The Agent Roster

Every Project Keeps
Its Own Roster

Create, task, and remove agents per project. Each row is live: an agent is idle, thinking, has changes ready to land, or holds a fork ready for review. Broken intermediate states stay invisible to teammates — in-flight work is never shared until the agent lands it.

Two ways to land

Simple projects auto-merge an agent's edits to peers continuously — no gate, like a trusted teammate. Weave projects give you a review-and-land gate: you look the work over, then promote it with a click.

Proof you can trust

An agent can attach evidence to its work — carried right on the change. You land work you can verify, not promises.

✓ compiled ✓ tests passed ✓ linters clean
Which agents work with it?
Any agent that can edit files can attach — the materialized folder plus a local commit socket is the whole interface. The deepest, turnkey integration today is Claude Code: live thinking status via lifecycle hooks, one-line commits, and attached test/lint evidence. Codex and local agents attach through the same folder and socket.
Parallel, Conflict-Free

Run a Whole Team of Agents
at the Same Time

Fork a project into as many parallel folders as you like and point a different agent at each. Five agents, five approaches, at the same time — none of them stepping on the others. Each works in its own private fork; when they land, their edits converge with no merge conflicts, held together by the same conflict-free sync (CRDT + braid) that keeps human peers in step. It's branch-and-merge for AI collaborators, without the merge pain.

Parallel agent forks converging conflict-free One project line forks into three agent forks that work in parallel, then converge back into the shared project history with no merge conflicts. project agent A agent B agent C one history

Forks are private to your device — each peer owns its experimental forks

We build SyncEngine this way

This isn't a demo. Much of SyncEngine — and the brief this very page was written from — was made by AI agents working exactly like this: each in its own fork, braiding its slice into one shared history, alongside the humans. The tool that lets AI and people collaborate as peers was built by AI and people collaborating as peers.

The agent works where your work already lives. Its changes flow to your peers the way anyone's do. Humans and agents, working the same project — on your machine, no middle.