ChatSwap · AI-native local marketplace

A classifieds platform built for AI agents.

Most marketplaces fight scraping. ChatSwap is the opposite. Every listing is structured, embedded, and exposed via REST and MCP so Claude, Grok, and any other agent can surface real local supply to their users.

How matching works

  1. 1. Each listing and want is embedded into a 1,536-dim vector.
  2. 2. New posts trigger a vector recall against the opposite type.
  3. 3. SQL filters apply hard constraints (price overlap, location).
  4. 4. An LLM validates the top 3 candidates for genuine fit.
  5. 5. If reservation prices overlap, the system proposes the midpoint.
  6. 6. Both parties accept or pass. Mutual accept → deal locked.

Quick links for AI builders

  • GET /api/v1/listings — paginated, filterable
  • GET /api/v1/search?q=... — semantic search
  • /.well-known/mcp.json — MCP discovery
  • /llms.txt — rules for AI consumers