Moonlighting

Invisible hands for your AI agent. A real Chrome, on its own hidden Windows desktop, driven by element references instead of coordinates — with credentials that never reach the model.

English · 中文

┌─ your desktop ─────────────────────────────┐
│  Hermes / Claude / Codex / any MCP client   │
└───────────────┬────────────────────────────┘
                │ MCP (stdio)  or  local HTTP
┌───────────────▼────────────────────────────┐
│  hidden desktop "HermesDesk"                │
│  ┌──────────────────────────────────────┐   │
│  │ real Chrome + persistent profile     │   │
│  │ (logins, wallet, extensions)         │   │
│  └──────────────────────────────────────┘   │
│  credential broker (DPAPI, origin-bound)    │
└─────────────────────────────────────────────┘

You never see a window flash. The agent never sees a password.

Why

Driving a browser as an agent usually goes wrong in three places:

Install

Requires Windows 10/11, Python 3.10+, Node 18+ (for the Playwright MCP child), Chrome.

Option A — pip (no paths to edit)

pip install moonlighting
moonlighting-setup --options          # where should browser data live?
moonlighting-setup --home "D:\moonlighting"
moonlighting                          # starts stream + broker + driver

Then wire it into your MCP client with no absolute paths at all:

{ "mcpServers": { "browser": { "command": "moonlighting-mcp" } } }

moonlighting-mcp, moonlighting-setup, moonlighting-session and moonlighting are console scripts installed by the package.

Option B — clone

git clone https://github.com/xinyuzjj/moonlighting.git
cd moonlighting
pip install -r requirements.txt

python scripts/setup.py --options          # where should browser data live?
python scripts/setup.py --home "D:\moonlighting"
python scripts/svc.py                      # starts stream + broker + driver

Either way, install_mcp.py writes the right config for you — it detects whether moonlighting-mcp is installed and falls back to absolute script paths if not:

python scripts/install_mcp.py                 # print configs with real paths
python scripts/install_mcp.py --write claude  # write it (backs up the original)
python scripts/install_mcp.py --write codex
python scripts/install_mcp.py --write all

svc.py / moonlighting prints a preview URL. Open it in the Hermes preview pane (or any browser) to watch and even click into the hidden desktop:

preview: http://127.0.0.1:8765/ctl

First run creates <home>/token.txt — every state-changing endpoint requires it.

Use it from an agent

Any MCP client (Claude Desktop, Claude Code, Codex CLI, Hermes)

scripts/mcp_server.py is a standard MCP stdio server exposing 27 tools (25 browser_* plus cred_list / cred_fill).

If you installed with pip, the whole config is one line — no paths:

{ "mcpServers": { "browser": { "command": "moonlighting-mcp" } } }

Cloned instead? Let install_mcp.py fill in the real paths:

python scripts/install_mcp.py --write all

Or write it by hand (Claude Desktop / Claude Code, mcpServers):

{
  "mcpServers": {
    "browser": {
      "command": "python",
      "args": ["C:\\path\\to\\moonlighting\\scripts\\mcp_server.py",
               "--cdp-endpoint", "http://127.0.0.1:18763"]
    }
  }
}

Codex CLI (~/.codex/config.toml — TOML, snake_case, and give it more startup time):

[mcp_servers.browser]
command = "python"
args = ["C:\\path\\to\\moonlighting\\scripts\\mcp_server.py", "--cdp-endpoint", "http://127.0.0.1:18763"]
startup_timeout_sec = 60
tool_timeout_sec = 120

Plain HTTP

TOKEN=$(python scripts/auth.py --print | sed 's/^token: //')

curl -H "X-Hermes-Token: $TOKEN" http://127.0.0.1:8897/snap        # whole page + refs
curl -H "X-Hermes-Token: $TOKEN" "http://127.0.0.1:8897/click?target=e42&element=Sign%20in"
curl -H "X-Hermes-Token: $TOKEN" "http://127.0.0.1:8899/list"      # saved credentials (no secrets)

Multiple agents, no interference

Each session gets its own desktop, profile, and ports — two agents cannot fight over tabs or invalidate each other's element refs.

python scripts/session.py new --name a1 --url https://example.com
python scripts/session.py list
python scripts/session.py stop --name a1

a1   cdp=18764  stream=8766  driver=8898  UP  http://127.0.0.1:8766/ctl

Running more than one instance

The defaults (HermesDesk, ports 8765/8897/8899/18763, one global pointer file) are meant for a single installation. If something is already running on the machine, give the second instance its own identity — all of it is environment variables:

export HERMES_BROWSER_HOME="D:\mb-agent"          # browser data (profiles, vault, token)
export HERMES_SETTINGS="D:\mb-agent\settings.json"  # don't touch the global pointer at all
export HERMES_DESK=MBAgentDesk
export HERMES_STREAM_PORT=18770 HERMES_BROKER_PORT=18771
export HERMES_DRIVER_PORT=18772 HERMES_CDP_PORT=18773

What the code does to keep them apart:

• The pointer file is never silently rewritten. setup.py --home refuses when settings.json already points somewhere else; --force overwrites after backing it up.
• Ports are claimed with an identity check. Services report home + desk on /health and /status; a port whose owner reports a different identity is treated as foreign and skipped, instead of being adopted ("it answered, so it must be mine").
• Environment survives the detach. Services are spawned through WMI so they outlive the agent, and WMI does not inherit your environment — so svc.py writes the isolation variables into a .cmd wrapper that starts the process. Without that, an isolated install silently falls back to the default home and ports.
• No legacy fallback once you've chosen a home. With HERMES_BROWSER_HOME set, profiles/ and vault/ resolve under it, never under a previous install's paths.
• The token exists before you need it. svc.py generates it at startup, so reading token.txt right after boot returns a value instead of an empty file.

MOONLIGHTING_NPX pins the npx binary when the MCP client starts you with a clean PATH.

Security model

• Passwords never enter the model context. The user types into a masked popup; the broker encrypts with Windows DPAPI, binds the secret to the page origin, and injects it. Tool results only contain {ok, verified, submitted, id_len, pwd_len}.
• Origin-bound. A credential saved for https://github.com is refused on any other origin.
• Token-gated. token.txt is generated on first run (mode 600). Read-only probes (/health, /status, /snapshot) stay open so the preview keeps working.
• Audited. Every injection appends one line to <home>/audit.jsonl: timestamp, action, origin, handle, result code. No plaintext, ever.
• No CAPTCHA evasion. This drives a real browser; it does not spoof fingerprints, solve puzzles, or defeat anti-bot systems.

Layout

scripts/
  mcp_server.py      standard MCP stdio server (browser_* + cred_*)
  browser_daemon.py  persistent driver: snapshot / click / type / tabs / eval
  desk_stream.py     live view of the hidden desktop (CDP screencast) + click/type back-channel
  vault_broker.py    credential broker: DPAPI, origin check, selector fill with read-back verify
  cred_dialog.py     masked popup the user types into
  login_flow.py      full login loop: popup → fill → verify → submit → retry on failure
  desk_chrome.py     launch / close Chrome on a named private desktop
  private_desktop.py Win32 desktop primitives (CreateDesktopW, WM_SETTEXT, PrintWindow)
  session.py         per-agent isolation (desktop + profile + ports)
  svc.py             start/stop the three services silently (no console windows)
  ports.py           port allocation: if the default is taken, pick a free one
  auth.py            token file + header helper
templates/           preview pages (/ctl interactive, /view read-only)
docs/SKILL.md        the long-form playbook: every pitfall we hit, with measurements

Platform

Windows only, by design: private desktops (CreateDesktopW), DPAPI, and silent process launch are Win32 primitives. The browser-control layer itself is portable — porting means replacing private_desktop.py and the DPAPI calls.

Agent loop (experimental, agent/)

scripts/ is the hands; agent/ is a thin closed loop on top of it — not packaged in the wheel, run it from a clone.

cd moonlighting/agent
python run.py ../tasks/github-login-replay.json     # steps + assertions + retries + events.jsonl

Measured: 42 KB snapshot → 6.0 KB (-86%), empty click caught 10/10, failed step retries 3× then screenshots and rolls back. There is no planner yet: a task is an explicit step list.

Acknowledgements

Standing on these projects:

Design ideas borrowed, no code vendored:

• browserbase/stagehand — observe() returning real selectors instead of pixels; that is why this project resolves elements by selector or ref and never by coordinates.
• browser-use/browser-use — the agent loop shape: snapshot, act, verify.
• web-infra-dev/midscene — the vision-based fallback idea for canvas-only UIs (deliberately not enabled by default; it is slower).

License

MIT. The Apache-2.0 licenses of playwright-mcp and Playwright apply to those components.