Your assistant describes your question and gets back posts that may already answer it, people who have said they can help with that kind of thing, and a ready-to-publish version of the question. It publishes nothing and tells no one, so your assistant can try it several times while it sharpens the question. Each match is marked strong (from inside your trust network) or weak (a loose match), and weak matches are offered as suggestions only.
Ask it like this
- “Does anyone know how to appeal a property tax assessment?”
- “Has anyone here worked out how to self-host a chat server?”
When another tool fits better
- Who should I meet?
You want people to get to know, not help with one question. - Can I trust this source?
You want to know whether a specific source or person is reliable.
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.
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
find_help- Title
- Who can help me with this? (free, and nobody is told)
- Description
Use this when the user says "who can help me with…", "does anyone know about…", or "who can I ask about…", or is otherwise stuck on a question — "who can help me with this?", "has anyone here worked this out?", "is this already answered?". It finds who or what on UpTrust can help with the question, WITHOUT publishing anything and without notifying anyone. Call it the moment the user is stuck, and call it speculatively: it is free, silent, and safe to call several times in a turn while you refine the question. It returns posts that may already answer the question, people who said they can help with this kind of thing, a ready-to-publish version of the ask, and a one-line `speakable` summary you can relay as-is instead of writing your own claims. Use who_should_principal_meet instead when the user wants people to get to know rather than help with a question, and resolve_entity then is_trusted_by_principal when they ask whether a specific source or person is reliable. Everything is viewer-relative to the authenticated user: the ordering is that user's own trust graph, not a global ranking. READ `strength` on every match before you present it — "strong" means a solid match from inside the user's trust network, while "weak" means a loose lexical match, or an author the user has no trust path to, and must be offered as a suggestion ("this might be related"), never as the authoritative answer. When nothing here helps, READ `nextStep`: it is present only on a fully empty result and says what to do about it in this session — usually offering to submit the introduction that would give the graph something to rank on, or publishing `suggestedAskText` as an ask so the question reaches people. Relay it as a suggestion; this tool never publishes or submits on its own.