The Elora Taurus Project · Explore our work
How external AI assistants can connect to Elora through delegated identities, OAuth and MCP, with a bounded synthetic demonstration and explicit permission checks.
Implemented demo scope
External assistants retain their own inference provider. Owner consent binds a distinct Agent and registered OAuth client to an explicitly delegated scope.
Read the integration overview ↗Governed connection
Short-lived access credentials, rotating refresh credentials and per-operation permission checks keep connection identity separate from permission to retrieve evidence. Suspension, expiry and relevant owner changes invalidate authority.
Inspect E110 in Engineering ↗Read-only foundation
Agent Provenance and Security previews show retained audit receipts with missing historical fields preserved. The previews do not establish a shared threat detector or automatic security investigation.
Explore provenanceThe public integration demonstration is scoped to synthetic, read-only evidence. Local tests cover the connection and permission boundaries; this guide does not establish acceptance by every MCP client or access to the wider Research collection.
Separate local prototypes explore staged task plans, retained task-ledger evidence and review holds. They remain synthetic development work. Bespoke analysis, report drafting, rendering and publication require their own governed integration and approval boundaries before availability can be claimed.
