Margin — agent waitlist Margin is building a way for people to give AI agents controlled access to bank and brokerage accounts. This endpoint records interest in the product. It does not open an account, connect a financial account, or authorize transactions. Before submitting - Get the human owner's explicit permission to join the waitlist using their email address. - Use the agent endpoint below. Human submissions use the Request access form. - Do not send passwords, API keys, account numbers, or financial records. Submit POST /_api/agent-waitlist Content-Type: application/json Use the same origin as this page. No API key is required. Required fields email: The consenting human owner's contact email, at most 254 characters. agent: The agent's name, from 1 to 120 characters. owner_consent: The boolean true, only after the owner has agreed. Optional field website: Leave this empty or omit it. It is a spam-prevention field. Example { "email": "owner@example.com", "agent": "Hermes", "owner_consent": true } Responses 200: {"ok":true} A duplicate email in the agent channel returns the same success response. 400: {"error":"..."} — check the email, agent name, consent, and field names. 413: {"error":"..."} — the request is over the 4096-byte limit. 415: {"error":"..."} — use Content-Type: application/json. 503: {"error":"..."} — saving failed; retry later. Do not add submitter_type or other fields. The server records the agent channel. Ordinary JSON and legacy SuperJSON {"json":{...}} requests are accepted. A legacy request receives the same JSON envelope in its response. Data recorded The normalized contact email, agent name, submission time, agent channel, and self-reported owner consent are saved privately. There is no public endpoint for reading waitlist submissions. The same email may join once through the human form and once through this agent endpoint. The channel records which endpoint received the submission. It is not proof of human identity, agent identity, or independently verified owner consent. Documents /agents — readable instructions /agents.txt — plaintext instructions on the published site /_api/agents — plaintext instructions in preview and on the published site