Connect your AI assistant

Gnomoncontext holds something the open web doesn’t: live, anonymized human demand — what real people are looking for and offering, by topic and place. Give your assistant the MCP server and it can search that demand, find the complementary people, and (with your user’s consent) file their intent to broker a real introduction.

MCP endpoint

https://gnomoncontext.com/api/mcp

Streamable HTTP. Reads are public (rate-limited); writes need a bearer token.

ChatGPT — as an app

Once it’s published, find Gnomoncontext in the ChatGPT App Directory (chatgpt.com/apps, or the apps menu in the composer) and turn it on — no setup needed.

ChatGPT — connect it yourself

To use it before it’s listed, or to test your own build:

1. Settings → Apps → Advanced settings → Developer mode (on).
2. Create app and paste the MCP endpoint above. It must be a public HTTPS URL — localhost can’t be reached by ChatGPT.
3. In a chat, open the Developer Mode tool and ask: “Search Gnomoncontext for language exchange in Geneva.”

Claude

Add a remote MCP server pointing at the same URL:

{
  "mcpServers": {
    "gnomoncontext": {
      "url": "https://gnomoncontext.com/api/mcp"
    }
  }
}

For local development with the stdio server, see .mcp.json in the repo.

Tools

• search_intents read
  Structured + fuzzy search over anonymized human needs/offers/dreams (theme, topic, country, city, direction, language).
• find_matches read
  Given a user's intent, return the complementary side (opposite-direction offers/seekers, or co-dreamers) to connect them with.
• demand read
  Aggregate demand for any topic/place: how many are looking, the seeking-vs-offering split, the hottest topics.
• search read
  ChatGPT-connector-compatible free-text search returning {id,title,url}.
• fetch read
  Fetch a full record by id from a search result.
• submit_intent write
  File one person's need/offer/dream/event (requires a bearer token; consent-gated).

Writing (submit_intent)

Filing a person’s intent requires a bearer token (Authorization: Bearer …) and is consent-gated. Contact details are never returned by reads; they are exchanged only when both sides opt in. Machine descriptor: /.well-known/mcp · plaintext: /llms.txt.

← Home · Privacy