API
The Kyvvu API is a FastAPI backend that sits at the centre of the platform. Every agent interaction, policy evaluation, incident, and audit report flows through it.
Local:
http://localhost:8000Production: Your installation.
Interactive docs (Swagger):
/docsOpenAPI spec:
/openapi.json
Two types of callers
The API has two distinct audiences, each with different authentication and a different set of endpoints.
Agents — API key auth
Agents call exactly two endpoints, authenticated with an API key in the Authorization header:
Authorization: Bearer KvKey-your-api-keyPOST
/api/v1/agents
Register or update the agent
POST
/api/v1/logs
Log one execution step
That's all an agent ever needs to do. Everything else — policy evaluation, incident creation, hash chaining — happens server-side in response to these two calls.
For Microsoft Copilot Studio agents, there is one additional endpoint that combines registration and logging in a single call:
POST
/api/v1/copilot/log
Log a Copilot Studio step (auto-registers agent on first call)
Humans & the dashboard — JWT auth
All other endpoints are for human operators and the dashboard. They require a JWT token obtained by logging in:
These endpoints cover the full management surface of the platform:
auth/
User accounts, login, API key management
agents/
View, filter, and deactivate registered agents (GET, DELETE)
logs/
Browse execution logs, retrieve task chains, validate hash integrity (GET)
policies/
Create, update, enable/disable, and delete compliance policies
policy-templates/
Apply or remove pre-built policy frameworks in one operation
incidents/
Review, resolve, ignore, and reactivate policy violations
reports/
Generate PDF or XML audit reports for any date range
events/
Immutable platform audit trail (read-only)
actions/
Configure automated responses to incidents (webhooks, alerts)
Authentication summary
AI agent
API key (KvKey-...)
POST /agents, POST /logs, POST /copilot/log
Human / dashboard
JWT token (email + password)
Everything else
API keys are scoped to a user account. An agent can only register itself and log its own steps — it cannot read other agents' data, access policies, or view incidents.
Health check
Returns overall status, uptime, system metrics, and database connectivity. No authentication required.
Further reading
For a full description of each module's data model and behaviour, see API Reference. For the interactive endpoint documentation, visit /docs on your running instance.
Last updated
