Managed context connectors
Connect knowledge, observability, delivery, and data sources through bounded read-only credentials.
Managed connectors turn provider records into tenant-scoped evidence without giving an agent ambient authority. Owners configure them in Source Readiness → Connector catalog. Hyperoru validates the provider before encrypting the credential, bounds every collection, and preserves provenance and freshness with the accepted evidence.
Available connectors
| Source | Required boundary | Evidence collected |
|---|---|---|
| Jira | Site and optional project keys | Work, ownership, remediation, releases |
| Confluence | Site and optional space IDs | Pages, decisions, ownership, runbooks |
| Notion | Explicitly shared pages and optional title filter | Documents, decisions, ownership |
| Google Drive | Connected account and optional shared-drive ID | File metadata and knowledge context |
| Sentry | Organization slug | Runtime issues and release context |
| Datadog | Restricted application key | Service definitions and runtime context |
| CircleCI | Project slug | Pipelines and delivery context |
| Jenkins | Controller with Job/Read | Jobs and latest build context |
| Airbyte | Optional workspace-ID allowlist | Data sources and lineage context |
Credential lifecycle
The setup dialog performs a live read-only request before saving. Credentials are encrypted and never returned. Owners can rotate, pause, resume, test, retry, or revoke a connection. Revocation deletes the credential envelope and stops collection, while already accepted evidence remains linked to its original provenance.
Duplicate handling
Each provider record receives a deterministic subject identity and a content hash. Replays update the same subject instead of publishing duplicate context. Scanner findings retain every source observation behind one canonical finding identity; similarity never silently merges findings.
Production canary
Use the dedicated customer-canary runner to prove the GitHub, cloud, Jira, Notion, and ZIP paths in the deployed environment. The generated artifact includes only identifiers, counts, states, and safe errors—never tokens or provider response bodies.