API & MCP

Read your verified model programmatically — a REST API for your systems, and an MCP server so an AI copilot can consume the governed context directly.

Overview

The API gives your own tools and AI copilots read access to the verified model — the same evidence-backed facts you see in the app, scoped to your organization. There are two ways in: a plain REST API, and an MCP server for AI assistants that speak the Model Context Protocol.

💡

This page explains what the API is for. For the endpoint-by-endpoint detail — parameters, response examples, a copyable curl and a console that runs the call with your own key — see the API Reference. The OpenAPI 3.1 document is at /api/v1/openapi.json.

Authentication

Create and manage API keys under API keys in your account. Keys are scoped to your organization and are stored hashed — the full key is shown once, at creation. Send it as a bearer token (Authorization: Bearer <key>) or the X-API-Key header.

⚠️

Treat an API key like a password. It grants read access to your organization’s models. Copy it when it is created; you cannot see it again, only revoke it and issue a new one.

REST endpoints

The REST API is read-only over your models, with a small ingest surface for AI-governance events. Each resource below is documented endpoint-by-endpoint, with live examples, in the API Reference:

Resource	What it returns
Assessments	The systems in your organization and their status.
Assessment behaviours	The verified behaviours of a system, with evidence and review status.
Governed context	The AI-ready context: verified facts, known unknowns, and the usage policy.
Changes	What changed between model baselines, including invalidated sign-offs.
AI events / exposure	Post AI-runtime events, or exposure findings, into your governance record.

The governed context

The context endpoint is purpose-built for feeding an AI copilot safely: it returns the verified behaviours it may rely on, the known-unknowns marked “do not assert”, and the read-only/advisory usage policy the model is served under. See Governance & AI for what each part means.

MCP for AI copilots

Daritas exposes a Model Context Protocol server, so an MCP-capable assistant can connect with an API key and call tools to list assessments, fetch a system’s behaviours, pull the governed context, and read what changed — all scoped to your organization, all grounded in the verified model.

💡

The point of consuming Daritas over MCP is that the copilot answers from a verified, cited model with explicit known-unknowns — not from an unaudited guess at your control logic.

Scoping & safety

Every key is scoped to a single organization; it can only read that org’s models.
The model API is read-only. The only writes are AI-governance events you choose to record.
Nothing in the API instructs a control action — it serves understanding and governance, consistent with the advisory boundary.