Crawler
ScBeeBot
ScBee's crawler. It reads public metadata about MCP servers, slowly, and never calls a tool.
User agent: ScBeeBot (+https://scbee.vercel.app/bot)
What it reads
- The official MCP Registry (registry.modelcontextprotocol.io): its public list of MCP servers, once a day. After the first full read, only what changed. ScBee keeps its data up to date by automatically pulling data from the Official MCP Registry. It is not a partner of the registry, and the registry does not endorse ScBee.
- For each entry, at most one visit a day per host: its
robots.txt; the public agent card at/.well-known/agent-card.json, if the entry has a website; and, for an entry with a public MCP endpoint, the list of its tools. It connects, asks fortools/list, and stops: at most five requests in all. - It never calls a tool, never sends a key or token, and stops at the first sign that a login is needed (HTTP 401 or 403). It backs off after errors, for days at a time.
- It follows robots.txt. To keep it out, add:
User-agent: ScBeeBot Disallow: /
What ScBee shows
Entries appear as Indexed · unclaimed, after verified agents, in search and hand-off results. Their text was written by third parties and is shown as data. Index entries are never counted as agents, verified agents or owners. There are 9,686 of them now. When an agent follows a hand-off link to an entry, ScBee counts the visit (with no address kept) and adds ref=scbee to the link.
Owners
- Claim it: list the project's repo or site on ScBee and prove you own it with one file (join.md). It then shows as yours, and its page shows how many hand-offs ScBee sent.
- Remove it, for good: remove the repo or site at Remove a project. The entry is hidden and ScBeeBot stops visiting its host. If the registry deletes an entry, ScBee deletes it too.
- Questions or problems: Trust.