Skip to content

Autopoetic Documentation

Autopoetic is an agentic AI platform. Agency is not a monolith here — it is a composition of three distinct, interoperable systems:

  • Inference — the reasoning engine. Our model, The Humanist, is tuned for a subjective Mixture of Perspectives architecture that co-evaluate every message and mathematically combine to expand and deepen its point of view.
  • Memory — the persistent, structured knowledge of the world.
  • Workflows — the orchestration and decision engine.

These systems are usable separately or together. Plug Autopoetic Memory into your existing agent stack to give it persistent, auditable knowledge. Call the Responses API to run Autopoetic inference — subjective, multi-perspective generation — from your own agents and tools. Or deploy the full Autopoetic stack, where inference, memory, and workflows operate as a coherent, self-sustaining loop.

Prefer nothing to assemble? Chat is that full stack as one easy-to-use interface — all three systems bundled seamlessly. Every message runs inference through the agent's perspectives, reads from and grows the project's memory, and the same conversations can be driven by your workflows when you need orchestration. See your first chat.

This is true agency: not just a model that answers questions, but a system that remembers, interprets, reasons, and decides — and every component is modular, auditable, and portable.

What you can do

  • Chat with an agent that has a point of view. Every message is evaluated by multiple internal perspectives — a cynic, an empiricist, a speculator, an empath — whose collective influence shapes the response. Watch each perspective's take stream live, see how they weighted the answer, and explore the "subjective shape" of every message in 3D. Start with what Autopoetic is.
  • Build a memory automatically. Every conversation teaches Autopoetic about the people, places, and things you discuss — entities, relationships, and a hash-chained event log you can explore as an interactive graph, complete with a time slider. Or plug that memory into your own agent stack as a persistent, auditable knowledge layer. See memory.
  • Orchestrate agents and decisions. Workflow templates compose steps that call out to your systems via webhooks, wait (indefinitely, for free) for your agents to respond, and gate work behind peer review whose confidence is computed by Autopoetic's decision engine. See workflows.
  • Use it programmatically. The Responses API generates agent responses for your code; the platform REST API manages conversations, memory, projects, and workflows as data; the MCP server drives the same from LLM clients. See the API overview.

Where to go next