Open Source

Your agents can't share files. That's insane.

P2P encrypted file sync for AI agents. Two commands, zero config.

Paste this into any AI agent

Works with Claude, ChatGPT, OpenClaw, Cursor, or any agent with terminal access.

Prompt

Set up botsync for P2P file sync. Run npx botsync init and give me the 5-word pairing code. I'll use it on the other machine to connect.

Your agent handles the install, config, and daemon setup. You just pass the code to the other side.

Requires Node.js 18+ on the machine · View on GitHub

Initialize on Machine A

Starts Syncthing, registers with the relay, gives you a pairing code.

npx botsync init

Join from Machine B

Pass the 5-word code. Peer discovery and encrypted pairing happen automatically.

npx botsync join north-crane-polar-quake-drift

Done — files sync in real-time

Everything in ~/sync/ flows between machines. Encrypted, P2P, survives reboots.

~/sync/shared/ ⇄ ~/sync/shared/

Requires Node.js 18+ · View on GitHub

Why botsync

The sync layer agents deserve

🔒

No Cloud

Direct peer-to-peer. Your files never touch a server. The relay only brokers connections — it never sees your data.

🛡️

Encrypted

TLS 1.3 in transit. Authenticated relay. One-time pairing codes. Your data stays yours.

⚡

Zero Config

No YAML. No env vars. No API keys. One command to start, one to join.

♾️

Survives Reboots

Set it up once. Runs as a background service. Files keep flowing after restarts.

🆓

Free & Open Source

MIT licensed. No vendor lock-in. Ever.

📡

Live Dashboard

See every connected device and last heartbeat. Open dashboard →

Under the Hood

What happens when you run botsync

No magic. Just well-composed open-source tools.

Architecture

✓ Syncthing handles P2P sync — battle-tested, 60k+ GitHub stars
✓ Relay brokers discovery only — never sees your files
✓ 5-word pairing codes (50-bit entropy, 10-min TTL, one-time use)
✓ Authenticated heartbeats — network secret generated at init, hashed on relay
✓ Background daemon — survives reboots, auto-restarts
✓ Works on macOS, Linux, WSL — anywhere Node.js runs

~/sync/
├── shared/ ← everyone sees everything
├── deliverables/ ← agents write, humans read
├── inbox/ ← humans drop files for agents
└── .botsync ← config (auto-generated)

How botsync compares

	botsync	scp / rsync	S3 / GCS	Syncthing (raw)	Git
Setup time	~30 seconds	SSH keys + paths	IAM + buckets + SDK	Device IDs + XML config	Repo + SSH + merge conflicts
Real-time sync	✓	✗ (manual)	✗ (poll or trigger)	✓	✗ (commit + push)
P2P (no cloud)	✓	✓	✗	✓	✗ (needs remote)
Agent-friendly	✓	Barely	SDK bloat	Config nightmare	Merge conflicts
Survives reboots	✓	✗	✓	✓	✗
Zero config	✓	✗	✗	✗	✗

Use Cases

Built for agent workflows

The filesystem is the universal interface. botsync makes it work across machines.

🤝

The Handoff

Agent A writes a report. Agent B on another machine picks it up and keeps going. No API, no queue — just files.

Agent A→report.md→Agent B

🐝

The Swarm

Multiple coding agents sharing a project folder across different machines. Collaborative development without Git gymnastics.

Agent 1⇄shared/⇄Agent 2

🌉

The Human Bridge

Agent writes to ~/sync/deliverables/ and it appears on your laptop. Review outputs without SSH or file transfers.

Agent→deliverables/→You

🔄

The Multi-Machine Agent

Same agent, persistent state across your laptop, server, and VPS. Context follows wherever the agent runs.

Laptop⇄Server⇄VPS

⛓️

The Tool Chain

Image generation writes to shared/. Website builder picks it up. Deployer ships it. All coordinated through the filesystem.

gen→build→deploy