# Persistent webhooks, HTTP callback queues, and an email inbox for AI agents

> AgentAddress is designed for AI agents that need to receive something after their current process exits.

An AgentAddress is one task-scoped resource with:
- a public, write-only HTTPS inbox for webhooks and callbacks;
- an inbound email address;
- a private, ordered event queue;
- a one-time read token used by a future agent run.

## Lifecycle

1. Create an AgentAddress with `POST https://agentaddress.onrender.com/api/v1/addresses`.
2. Save the complete response before the current run ends. The read token cannot be recovered.
3. Give `endpoints.inbox_url` or `address.email` to the external responder.
4. The current process may exit; the AgentAddress continues receiving.
5. Return later and read `endpoints.events_url` using `credentials.read_token`.
6. Process events in `sequence` order, persist `next_cursor`, and acknowledge handled events.

## Machine entry points

- [Agent-readable index](https://agentaddress.onrender.com/llms.txt)
- [Complete agent guide](https://agentaddress.onrender.com/llms-full.txt)
- [OpenAPI 3.1 contract](https://agentaddress.onrender.com/openapi.json)
- [Capability discovery](https://agentaddress.onrender.com/.well-known/agentaddress.json)
- [Installable Agent Skill](https://agentaddress.onrender.com/skill.md)
- [MCP endpoint](https://agentaddress.onrender.com/api/mcp)