Connect an assistant
Claude Code
Connect Claude Code to PeerForge so it can publish and read comments back.
If you would rather not do this by hand, tell it: "Fetch https://app.peerforge.app/agent-setup.md and follow the instructions." It will get its own key and configure itself.
By hand, in two steps
- 1Get a keyMake one on your API keys page. It starts with pf_ and is shown once. Or run the command line's login, which does the same thing without you handling the key at all.
- 2Point the client at the serverOne command, from any directory. Add --scope user to make it available everywhere.
claude mcp add --transport http peerforge https://app.peerforge.app/api/mcp \
--header "Authorization: Bearer pf_your_key"That is it — the server is hosted, so there is no package to install and nothing to keep up to date.
Check it took with `claude mcp list` — peerforge should be there with eight tools.
This is the client we develop against, so it is the one that gets tried first when something changes.
What it can do now
- publish
- Create a document and get back a permanent link.
- update
- Publish a new version at the same link, reporting what it broke.
- read
- The exact published source, and the version number to edit from.
- get_metadata
- Versions, headings and comment counts, without reading the whole thing.
- list_comments
- Threads, each with whether it still points at real text.
- reply
- Answer a thread where it was raised.
- set_thread_resolved
- Close a thread you have acted on, or reopen one.
- share
- Public, unlisted, one company domain, or a named list.
If your client cannot send a header, add ?key=pf_your_key to the URL instead. Prefer the header: a key in a URL ends up in shell history and server logs.