API
Reasoning primitives you compose around your own data.

No fixed workflow. The API exposes individual reasoning capabilities that you compose around your own data, applications and operating logic. Every call operates on sessions, beliefs, belief states, dissonance or reasoning processes.
The catalog
The API catalog lists every primitive, grouped by capability: sessions, extraction, beliefs, counterfactual reasoning, dissonance, the belief graph and history. Each entry shows its inputs and structured outputs. Each one runs from the playground against a session you pick.
Use it to:
- discover the reasoning primitives;
- read the inputs and outputs of each one;
- test a single operation on a live session;
- chain operations into your own workflow;
- wire belief states into agents, applications and internal systems.
Answers come back as beliefs, with confidence, evidence and conflicts attached.

Sessions first
A session is the container. It holds one belief state, its evidence and its history. It can model a person, a team, a customer account, a project, an investigation or an agent. Send evidence in. Beliefs and structure come out. Ask a question and the reply cites the beliefs it rests on, each with its confidence and conflicts.
How it connects
- Direct API calls.
- MCP access for agents and applications.
- Event listeners and subscriptions.
- Workflow triggers.
Inputs can be a fixed corpus or a live stream: documents, messages, databases, system events, model and agent outputs, tool activity and human feedback.