Let any MCP-compatible agent work in your instance, as a named person, with an expiry date.
What it does
- Any MCP client can use your instance as a tool source.
- An agent acts as the person who made its token, and can reach nothing more.
- Read-only tokens are offered no write tools at all.
- Every token is scoped to one nexus and expires.
- Anyone makes their own; an administrator can revoke any of them.
- Installed plugins add their own tools automatically.
An agent session, recorded
A small terminal client of our own, connected over MCP with a token, finds a page, reads it, looks up a figure in a requirements collection, and writes a section citing it. The calls are scripted so the recording can be repeated, and every result on screen is what the server answered. The section goes into the page's draft, so a person sees it arrive and decides whether to publish it.
A token borrows one person's permissions
A token carries the exact permissions of the person who created it, and nothing more. It stops working when they revoke it or when it expires. Name, access mode, nexus scope and expiry are all set when the token is made.
Minting and revoking are different jobs
Anyone can mint a token for themselves. An instance administrator sees the full list and can revoke any of them, but cannot create a token on someone else's behalf.
What an agent can do
Search and read pages, create and edit them, work with spaces, labels, comments and page properties. Installed plugins add their own: an entitled Requirements collection exposes its items, traces and compliance report to the same agent, through the same endpoint, with the same permission rules.