Give your assistant a list you already have, such as papers, packages, newsletters or people, and it puts them in order of how much the people you trust rely on each one, with a one-line reason. It gets an order, never scores. Items your network has no signal on are left out, and the reason says which. It can also rank the same list as several communities see it, side by side, which shows what holds its place across them.
Ask it like this
- “Which of these five papers should I read first?”
- “Rank these libraries by what is safest to depend on.”
When another tool fits better
- Can I trust this source?
You want a yes or no on a single item. - Whose judgment do the people I trust rely on for this topic?
You have no list yet and want candidates found for you.
Before you use it
- It needs the Read your trust signal permission, which UpTrust asks you to approve when you connect.
- It only reads. It changes nothing on UpTrust.
- In Claude, connect
https://mcp.uptrusting.com/mcp. In ChatGPT, connecthttps://mcp.uptrusting.com/chatgpt. How to connect. - It can also answer from another trust graph than yours, such as one built from academic papers; see which communities it can answer from.
What your assistant is told
Your assistant decides when to use this tool from the title and description UpTrust’s server gives it. Here they are exactly as the server sends them.
- Tool name
rank_for_principal- Title
- Rank a list you already have by trust
- Description
Use this when the user already has several candidates — papers, repos, packages, sources, people, vendors, options — and asks which to trust or use first: "which of these is any good?", "rank these", "which should I read first?" (when a list is present), "sort these by what's safest to depend on". It returns the candidates' IDs ordered from most to least trusted by the user's own trust network (the "principal" in the name is the signed-in user), or by a named community when you route the call to one, with a one-line `rationale`. You get an ordering, never raw scores. Unrecognized IDs, and ones the user has no signal on, are dropped and accounted for in `rationale`. It compares only the candidates you pass in. Use top_trusted_for_principal instead when there is no list yet (it discovers candidates), and is_trusted_by_principal for a single go/no-go on one item. Pass each candidate as `{ id, kind }` in this issuer's namespace — the pair resolve_entity returns; call it first if you are holding URLs, titles or names. A post is ordered by the user's trust in its author, putting people and content on one scale. WHEN MORE THAN ONE COMMUNITY HAS A CLAIM ON THE ANSWER, PASS `compare_perspectives` instead of calling this once per community: it ranks the same candidates as each one and returns the orderings side by side, plus which items held their position and which moved. An item that keeps its place across communities that disagree about everything else is a stronger signal than any single ordering, and only the comparison shows it. PRESENTATION: lead with the recommendation and the one reason it holds. The basis, the caveats and the dropped IDs go after it, or on request — a user choosing between options wants the pick, and an answer that opens with epistemics reads as a hedge. Never open by reporting which graph or corpus produced the answer; that is routing, not an answer.