Integrations & APIs

Available today

• Agent enforcement: before acting, your agent or gateway POSTs to /api/runtime/agents/{projectId}/authorize with the X-Agent-Enforcement-Key header, and reports what it did to the traces endpoint. The decision checks the agent's allowed and prohibited actions, per-tool permissions, external authorisation and spend limit. Keys are created and revoked under Enforcement keys in Agent & MCP Governance. Calls are refused while the workspace is read-only.
• Monitoring signals: POST to /api/runtime/signals with the X-Monitoring-Key header. Keys are created when you add a signal source in Monitoring. Each source accepts up to 120 signals a minute; more receive HTTP 429.
• Evaluation results: your runner POSTs results to /api/evaluations/results with an HMAC-SHA256 signature in the X-AIGovernr-Signature header. The signing key for each project is set up by AIGovernr; it cannot be created in the app.

Enforcement and monitoring keys are shown once when created. Store them in a secrets manager. Revoke a leaked enforcement key under Enforcement keys. A monitoring key cannot be revoked in the app yet.

Evaluation CI template

CI config in Evaluations generates a GitHub Actions workflow for an approved specification. Treat it as a starting point: its submit step calls an aigovernr-evaluator command that is not published, and it does not include your AIGovernr address or project ID. Replace that step with your own signed submission.

Integrations hub

Project → Integrations lists connectors such as GitHub, Jira, ServiceNow, Slack and Teams, and stores their credentials and settings. This deployment does not include the remote network adapters, so nothing is synced from those services yet. Check stored credential confirms only that a credential is saved, not that the service can be reached.

Planned

• A developer CLI and CI checks.
• A general public REST API. The application's own endpoints are not a supported public API.
• Repository scanning on push and pull request.

Check the capability matrix on the Trust Center for current status before relying on a planned feature.