Skip to Content

hosted mcp server

the hosted server runs at https://api.crawlbrulee.com/mcp and speaks streamable http. there’s nothing to install, and you sign in through your browser instead of handling an api key. it has the same seven tools as the local server; see mcp server.

sign in

the first time your app connects, it opens a crawlbrulee sign-in page in your browser. sign in, pick the organization to connect, and approve. if you don’t have an account yet, you can sign up  on the way; the free plan needs no credit card.

connect your app

Claude (claude.ai and Claude Desktop): open Settings → Connectors, add a custom connector with the url, and leave the OAuth client ID and secret empty.

Claude Code:

claude mcp add --transport http crawlbrulee https://api.crawlbrulee.com/mcp

then run /mcp, pick crawlbrulee and sign in.

Cursor: add to ~/.cursor/mcp.json (or .cursor/mcp.json in a project). Cursor asks you to sign in when it connects:

{ "mcpServers": { "crawlbrulee": { "url": "https://api.crawlbrulee.com/mcp" } } }

Codex: add to ~/.codex/config.toml, then run codex mcp login crawlbrulee:

[mcp_servers.crawlbrulee] url = "https://api.crawlbrulee.com/mcp"

Gemini CLI: add to ~/.gemini/settings.json, then run /mcp auth crawlbrulee:

{ "mcpServers": { "crawlbrulee": { "httpUrl": "https://api.crawlbrulee.com/mcp" } } }

other apps: add the url as a remote (streamable http) mcp server. the app finds the sign-in on its own.

your connections

each app you connect gets its own key in the organization you picked. it shows in the dashboard under tokens, marked “connected app”, next to your other keys. delete it to disconnect the app; the app asks you to sign in again on its next call. a free organization can have 25 connected apps at a time, a paid one 200.

an app that can’t sign in can send an api key instead, in the Authorization: Bearer cwbl_... header. add the header to the app’s config yourself; don’t paste the key into the chat.

what the hosted server leaves out

the hosted tools take the same input as the local ones, except:

  • no headers and cookies, no extract.raw_html, no screenshot viewport or actions, and no webhook on scrape_async. a call that sends one gets a validation_error. need them? use the local mcp server, the cli or an sdk.
  • map returns 200 urls a page unless you set limit.
  • scrape stops after about 50 seconds with request_timeout, before most chat apps give up on a call. use scrape_async for slow pages.

when a call fails

a failed call comes back as a tool result whose text is the api’s error object, as json, for example {"name":"validation_error","message":"..."}. see errors for every name.