Install CLI + Skill

CLI + Skill is the recommended integration for terminal-capable coding agents.
It gives your agent the commands and instructions to find tools, run requests,
and check the results.

Set up in your AI agent

Paste this into Codex, Claude Code, Cursor, or another coding agent:

Set up Scrappycoco: https://scrappycoco.ai/setup.md

The agent checks its own environment and sets up CLI + Skill. If Node/npm is
missing, the installer downloads a checksum-verified private runtime under your
user account. It does not require administrator access or change system Node or
shell startup files. Sign in in your browser when prompted.

The managed launcher is ~/.local/share/scrappycoco/bin/scrappycoco on macOS/Linux
and %LOCALAPPDATA%\Scrappycoco\bin\scrappycoco.cmd on Windows. Use that full
path in the current agent session; restarting the terminal is unnecessary.

Clients without terminal execution can use the remote MCP fallback. Setup will
explain any client limitation rather than claim the installation succeeded.

Direct terminal setup

If your terminal already has Node.js 20+ and npm, you can still run:

npx --yes @scrappycoco/cli@latest setup

The command:

1. Opens browser sign-in or account creation.
2. Stores the OAuth credential in the OS keychain when available, with a
   permission-restricted local fallback.
3. Installs the hash-verified Scrappycoco skill.
4. Verifies authentication and the live scraper catalog.

Do not treat a successful browser callback as completed setup. Wait for the
terminal to confirm credential storage and catalog verification.

Load the skill

Using the same working CLI invocation from setup, verify the installation
(replace scrappycoco with the full launcher path if needed):

scrappycoco auth status --json
scrappycoco skill status --json
scrappycoco catalog list --available --json

Read the SKILL.md at skill.path in the setup result, or target from
skill status, in the current conversation and apply its instructions now.
The CLI is ready immediately. Automatic skill discovery for future sessions
depends on the client and does not block current CLI use.

The CLI checks for skill updates daily. To check immediately:

scrappycoco skill update --json

Read the updated skill in the current conversation and continue.

Remote or headless terminal

When the browser cannot reach the CLI's loopback callback, keep the process
running and use:

scrappycoco setup --no-browser --manual-callback

Open the printed URL and paste the browser's final callback URL into the same
terminal. The URL contains a short-lived authorization code. Never paste it
into chat, logs, or a saved file. If the process exits or times out, start a new
setup command and use its new URL.

Noninteractive environments

Do not run browser OAuth in CI. Create a workspace API key in the Scrappycoco
dashboard and expose it to the process as SCRAPPYCOCO_API_KEY.

Next, use the CLI or choose another integration.