Connect with MCP

Use MCP when the client should expose Scrappycoco as native tools. The remote
Streamable HTTP endpoint is:

https://api.scrappycoco.ai/mcp

The server publishes OAuth protected-resource metadata and requires a token
scoped to the MCP resource.

Codex plugin

codex plugin marketplace add https://github.com/Albert-Tam/scrappycoco-integrations
codex plugin add scrappycoco@scrappycoco

Trust and load the plugin, then complete OAuth when the client asks. Confirm
that list_scrapers is available before treating the integration as complete.

Claude Code plugin

claude plugin marketplace add Albert-Tam/scrappycoco-integrations
claude plugin install scrappycoco@scrappycoco

For an installation made during a running session, use /reload-plugins, then
open /mcp, select scrappycoco, and authenticate. Plugin installation does
not launch MCP OAuth automatically.

Other MCP clients

Add https://api.scrappycoco.ai/mcp as a remote Streamable HTTP server. Follow
the client's trust and OAuth flow; do not place an API key in the MCP URL.

Recommended tool sequence

1. Call list_scrapers to read the live catalog.
2. Call get_scraper for the selected source.capability.
3. Build input only from the returned schema.
4. Use run_scraper when the configuration is known.
5. Use the discovery lifecycle when provider or native options are uncertain.
6. Inspect records, native outputs, attempts, cost, errors, and truncation.

The server does not invoke a Scrappycoco-owned LLM. The connected AI client
interprets the user's goal and judges candidate evidence.