{"version":1,"resources":["https://agentcash.honcho.dev/api/honcho/messages","https://agentcash.honcho.dev/api/honcho/messages/list","https://agentcash.honcho.dev/api/honcho/chat","https://agentcash.honcho.dev/api/honcho/representation","https://agentcash.honcho.dev/api/honcho/card","https://agentcash.honcho.dev/api/honcho/card/set","https://agentcash.honcho.dev/api/honcho/context","https://agentcash.honcho.dev/api/honcho/peers","https://agentcash.honcho.dev/api/honcho/peers/sessions","https://agentcash.honcho.dev/api/honcho/sessions","https://agentcash.honcho.dev/api/honcho/sessions/peers","https://agentcash.honcho.dev/api/honcho/sessions/peers/add","https://agentcash.honcho.dev/api/honcho/sessions/peers/remove","https://agentcash.honcho.dev/api/honcho/sessions/delete","https://agentcash.honcho.dev/api/honcho/sessions/summaries","https://agentcash.honcho.dev/api/honcho/search"],"mppResources":["https://agentcash.honcho.dev/api/honcho/messages","https://agentcash.honcho.dev/api/honcho/chat"],"description":"Persistent memory for AI agents. Use this to remember users across conversations, store and recall what was said, and build long-term context. Store messages, query memory in natural language, and retrieve learned representations and peer cards. Pay per request — no auth, no subscriptions. Powered by Honcho.","instructions":"# Honcho Memory API\n\nPay-per-call HTTP gateway over [Honcho](https://honcho.dev): persistent\nconversational memory for AI agents. Each wallet owns a private workspace,\nauto-derived from your SIWX signature (free endpoints) or payer address\n(paid endpoints) — never pass a workspace field. Use via the AgentCash MCP\nor CLI: `agentcash.discover` → `agentcash.checkSchema` → `agentcash.fetch`.\n\n## Recommended flow\n\nRecord every turn, query when personalization helps. Sessions and peers\nauto-provision on first reference.\n\n**1. Record the turn — PAID** (`POST /api/honcho/messages`)\n\n```json\n{\n  \"session\": \"<unique-id>\",\n  \"messages\": [\n    { \"peer\": \"<user-id>\",  \"content\": \"<user message>\" },\n    { \"peer\": \"Assistant\",  \"content\": \"<your response>\" }\n  ]\n}\n```\n\n$2 / 1M input tokens (estimated `ceil(total_chars / 4)`), $0.001 floor,\n$5 ceiling. Short turns (≲2,000 chars) settle at the floor. Always call\nthis after responding so Honcho can learn from the exchange. Store the\n`session` ID and reuse it for the rest of the conversation.\n\n**2. Query memory — PAID** (`POST /api/honcho/chat`)\n\n```json\n{\n  \"peer\": \"Assistant\",\n  \"query\": \"What communication style does this user prefer?\",\n  \"target\": \"<user-id>\",\n  \"session\": \"<session-id>\",\n  \"reasoning_level\": \"low\"\n}\n```\n\nCalls Honcho's reasoning system, grounded in everything it has learned\nacross the peer's conversations. Takes a few seconds — use it when\npersonalization would genuinely improve your response.\n\nGood queries:\n- \"What communication style does this user prefer?\"\n- \"How formal or casual should I be?\"\n- \"What is the user really asking for beyond their explicit question?\"\n- \"What emotional state might the user be in right now?\"\n\n| reasoning_level | Price  | Notes                          |\n|-----------------|--------|--------------------------------|\n| minimal         | $0.001 | basic, instant                 |\n| low (default)   | $0.01  | efficient synthesis, instant   |\n| medium          | $0.05  | steerable, fast                |\n| high            | $0.10  | deep synthesis, async          |\n| max             | $0.50  | research-grade, async          |\n\nUse the lowest level that fits.\n\n## Endpoints\n\nAll POST, JSON in/out. Full schema at `/openapi.json`,\ndiscovery at `/.well-known/x402`.\n\n| Endpoint                            | Auth | Use                                                          |\n|-------------------------------------|------|--------------------------------------------------------------|\n| `/api/honcho/messages`              | Paid | Append messages; auto-provisions session + peers.            |\n| `/api/honcho/chat`                  | Paid | Natural-language query over a peer's memory.                 |\n| `/api/honcho/messages/list`         | SIWX | Read session message history (paginated).                    |\n| `/api/honcho/peers`                 | SIWX | List peers in workspace.                                     |\n| `/api/honcho/peers/sessions`        | SIWX | List sessions a peer is in (paginated).                      |\n| `/api/honcho/representation`        | SIWX | Markdown representation of a peer.                           |\n| `/api/honcho/card`                  | SIWX | Compact biographical facts about a peer.                     |\n| `/api/honcho/card/set`              | SIWX | Manually set/correct a peer card.                            |\n| `/api/honcho/sessions`              | SIWX | List sessions (paginated).                                   |\n| `/api/honcho/sessions/peers`        | SIWX | List peers in a session.                                     |\n| `/api/honcho/sessions/peers/add`    | SIWX | Add peers to a session.                                      |\n| `/api/honcho/sessions/peers/remove` | SIWX | Remove peers from a session.                                 |\n| `/api/honcho/sessions/delete`       | SIWX | Delete a session and its messages.                           |\n| `/api/honcho/sessions/summaries`    | SIWX | Short + long session summaries.                              |\n| `/api/honcho/context`               | SIWX | LLM-ready context: recent messages + summary + rep/card.     |\n| `/api/honcho/search`                | SIWX | Semantic search; scope with `session` or `peer`.              |\n\n## Auto-provisioning\n\nYou don't need explicit \"create\" calls. `/api/honcho/messages` will create\nthe session, create each authoring peer, and add them to the session. Other\nendpoints (`/sessions/peers/add`, `/chat`, etc.) auto-create on first\nreference. Use `/sessions/peers/add` and `/sessions/peers/remove` directly\nwhen managing membership without sending messages.\n\n## Auth\n\nPaid endpoints use HTTP 402 micropayments (Base mainnet, Solana, or Tempo);\nworkspace = payer wallet. Free endpoints require a SIWX signature in the\n`X-SIWX` header; workspace = signing wallet. Memory is private per wallet —\nno cross-wallet access.\n"}