coding-bridge 2026.7.27.1

Node daemon for AceDataCloud Coding Bridge — run Claude Code on your own machine and drive it from the web.

Coding Bridge

Run Claude Code or OpenAI Codex on your own machine, and drive it from somewhere else — the AceDataCloud studio in a browser, your phone, or a WeChat / Telegram chat.

coding-bridge is a small outbound-only daemon. You pair it once with your Ace account, and after that your machine shows up as a remote-controllable device. Your code, files, shell, and tool approvals never leave your hardware — the relay only carries JSON messages.

Install

Requires Python 3.10+ and a signed-in Claude Code CLI. To use the Codex provider instead, install the Codex CLI and run codex login.

	Command
pipx (recommended)	pipx install coding-bridge
pip	pip install coding-bridge
Homebrew (macOS / Linux)	brew tap acedatacloud/tap && brew install coding-bridge
Scoop (Windows)	scoop bucket add acedata https://github.com/AceDataCloud/scoop-bucket scoop install coding-bridge
uv	uv tool install coding-bridge

On Windows without Scoop or pipx, use the Python launcher: py -m pip install --user --upgrade coding-bridge.

Check it landed:

coding-bridge status

Setup

1. Get a pair code. On the machine you want to control:

coding-bridge pair

It prints a short pair code, a direct link, and — with the qr extra — a QR code.

2. Claim it in the browser. Open the link the CLI printed (it fills the code in for you), or go to studio.acedata.cloud/coding-bridge, click Pair device, and type the code.

Sign in with your Ace account if you aren't already. Once claimed, the daemon stores a node token at ~/.ace-bridge/credentials.json (mode 0600) — you never pair this machine again.

3. Run it. Foreground, to try it out:

coding-bridge up          # pairs if needed, then serves sessions until Ctrl-C

Or as a background service that survives logout and reboot — this is what you want day to day:

coding-bridge service install    # registers, starts now, and enables at login

On a Homebrew install, brew services start coding-bridge starts the daemon through Homebrew's own service manager instead — pick one, not both. Either way, pair first (step 1): the service runs coding-bridge run, which exits non-zero when unpaired.

The node should now show online on the device page, and you can start a session from the browser or your phone.

coding-bridge service status              # ask the OS service manager
tail -f ~/.ace-bridge/logs/agent.log      # look for "registered with bridge"

Don't run coding-bridge up in a terminal while the service is active — two daemons sharing one node token would fight over the connection, so the second refuses to start.

More on the service (all subcommands, where the unit file lands, troubleshooting): docs/service.md.

How it works

The daemon opens one outbound WebSocket to the relay and waits. The browser connects to the same relay with your Ace JWT. The relay matches the two and forwards messages — it never sees your files and never runs anything.

┌────────────┐   wss (Ace JWT)   ┌───────────────┐   wss (node token)   ┌──────────────────────┐
│  Browser   │ ───────────────►  │ coding-bridge │ ───────────────────► │   coding-bridge      │
│  / phone   │ ◄───────────────  │    (relay)    │ ◄─────────────────── │   (this daemon)      │
└────────────┘                   └───────────────┘                      │    └─ Claude Code    │
                                                                        └──────────────────────┘
                                                                              runs on YOUR machine

That shape is the whole security model:

• Execution is local. Your repo, shell, and MCP servers only ever touch this machine.
• Outbound only. No listening ports, so nothing to expose or firewall.
• Per-tool approval — in the modes that ask. In default mode every tool the agent wants to run is relayed to you as a permission request; nothing runs until you allow it, and a timeout denies by default. The looser modes are looser on purpose: acceptEdits auto-accepts edits, and bypassPermissions runs everything unattended. Codex never asks interactively at all — codex exec maps the permission mode to a sandbox policy instead (plan → read-only, default/acceptEdits → workspace-write, bypassPermissions → danger-full-access). Pick the mode accordingly.
• Scoped, revocable token. Pairing yields a node token tied to your Ace account. coding-bridge logout removes it locally; the relay can revoke it.
• Your provider auth stays put. The agent drives your already-signed-in local Claude Code / Codex CLI.

Two local capabilities run inside that same boundary: fs.list lets the paired browser list directories so you can pick a working directory, and pasted images are decoded to <cwd>/.tmp/images/. Both act with the OS permissions of the account running the daemon.

The node is also the source of truth for what the UI offers: it answers capabilities.get with its providers, models, effort tiers, and permission modes (capabilities.py), so adding a model is a node-side change, never a web rebuild. The wire protocol is documented in protocol.py.

Drive it from WeChat or Telegram

Besides the browser, the same daemon can be driven from a chat account you already use. Message it, and each message runs one Claude Code / Codex turn whose reply lands back in the chat — handy for kicking off and steering tasks from your phone with nothing to install on it.

coding-bridge channels init      # writes ~/.ace-bridge/channels.toml
coding-bridge channels portal    # pick admins/groups in a local web UI
coding-bridge channels doctor    # validate config + tokens
coding-bridge channels start     # connect and serve

Both channels connect outbound only and keep the same local trust boundary. Every instance is disabled until you explicitly opt in, and a trigger prefix, sender allowlist, rate limit, and dedup all run before any provider turn.

Full setup — gateway config, Telegram bot token, the approval portal, running it as a service: docs/channels.md.

Reference

Commands:

Command	What it does
up	Pair if needed, then run (default if no command given)
pair	Pair this machine and exit
run	Run using stored credentials (errors if not paired)
status	Show configuration and whether this machine is paired
logout	Remove stored credentials
service	Manage the daemon as a background OS service — docs
channels	Drive the daemon from WeChat / Telegram — docs

Run flags (up / run): --model, --cwd, --permission-timeout, --claude-path, --codex-path, --copilot-path. Global flags: --bridge-url, --name, --config-dir.

Environment (see .env.example) — CLI flags win:

Variable	Default	Meaning
CODING_BRIDGE_URL	https://coding-bridge.acedata.cloud	Relay base URL
CODING_BRIDGE_NODE_NAME	hostname	Display name for this node
CODING_BRIDGE_CONFIG_DIR	~/.ace-bridge	Credential storage directory
CODING_BRIDGE_HEARTBEAT_INTERVAL	15	Heartbeat seconds
CODING_BRIDGE_PERMISSION_TIMEOUT	1800	Permission wait seconds (0 = forever)
CODING_BRIDGE_MODEL	—	Default model
CODING_BRIDGE_CLAIM_URL	https://studio.acedata.cloud/coding-bridge?code={code}	Pairing claim link template

Optional extras: coding-bridge[qr] for the ASCII-QR pairing helper; coding-bridge[wechat] / [telegram] are intent markers (they pull no extra wheels).

Development

pip install -e ".[dev]"
ruff check .
pytest -q

Attribution & license

Licensed under AGPL-3.0-or-later (see LICENSE).

The remote permission-relay design — forwarding a coding agent's tool-approval decision to a remote approver — was inspired by VibeBridge (GPL-3.0). This is an independent implementation built on the public claude-agent-sdk; no VibeBridge source code is included. See NOTICE.