An MCP server that gives AI assistants direct, cited access to the Canton Network's public record: improvement proposals, documentation, forum threads, mailing lists, whitepapers, grant proposals, releases and on-chain governance.
Hosted and free to use. No API key, no signup, read only.
https://ccpedia.xyz/mcp
Ask any assistant a specific Canton question today and you get one of three answers: a refusal, a generic blockchain answer that is not about Canton, or a confident invention. Everything needed to answer properly is public. It is just scattered across a dozen sites that no model was trained on, and none of them can be called by an assistant.
CCPEDIA indexes those sources continuously and exposes them as tools. An assistant can look up the actual text of a CIP, read what the forum concluded about it, check whether a vote passed, and cite all of it.
The full live inventory of what is indexed, how many records each source holds and when each was last synced is published here:
https://ccpedia.xyz/api/v1/sources
Those numbers live only there. This README does not repeat them on purpose: a count copied into a readme is out of date by the following week.
claude mcp add --transport http ccpedia https://ccpedia.xyz/mcp{
"mcpServers": {
"ccpedia": {
"type": "http",
"url": "https://ccpedia.xyz/mcp"
}
}
}Ready made files for each client are in examples/.
Streamable HTTP transport, protocol version 2024-11-05. Standard MCP
initialize then tools/list. Nothing custom.
Questions that used to need someone who had been reading the forum for a year:
- What does CIP-0042 actually say, and what was the objection to it
- Which CIPs are approved but not yet final
- What changed in the last Splice release that affects validators
- Has anyone proposed something like X before
- What did the Foundation decide about Y, and where is that written down
- Which grant proposals are live for a vote right now
See TOOLS.md for the full catalogue, grouped and with arguments.
It is generated from the running server rather than written by hand, so it cannot fall out of step with what the server actually offers.
This repository holds the documentation, client configuration, issue tracker and discussions for the hosted service.
It does not contain the server code, and it is worth saying why plainly. The
server is tied to the pipeline that builds the index and to the database it
reads. Publishing the request handler on its own would hand you something that
answers nothing. Separating them properly is real work, not a git push, so it
is on the list rather than in this commit.
If you want to run your own instance today, the same corpus is available over a
plain REST API, documented at https://ccpedia.xyz/api/v1.
- Discussions for questions, ideas, and "could it also do X"
- Issues for bugs, wrong answers, and corrections to indexed data
A wrong or missing answer is the most useful thing you can report. Include the question you asked and the tool the assistant called, if you can see it.
- ccpedia.xyz the site
- ccpedia.xyz/ask the same corpus as a chat agent
- ccpedia.xyz/api/v1 REST API for non-agent clients
Everything in this repository is MIT licensed. Indexed material stays under the license of whoever wrote it, and CCPEDIA links back to every source.