| Agent |
A program, often an AI assistant, that reads or writes data through InterLock. |
| Identity |
How InterLock knows an agent; every request is made by one. |
| API key |
The secret an identity authenticates with. Stored only as a peppered hash. |
| Data source |
One system agents reach through InterLock, such as a database, bucket or workspace. |
| Source ID |
The name agents use for a source: the database name on the PostgreSQL wire, source_id over MCP. |
| Connector |
The code that knows how to talk to one kind of source. Must be active before sources can use it. |
| Source role |
A named set of allow and deny statements on one source. |
| Statement |
One rule in a source role: effect, action, resource type, resource pattern, conditions. |
| Grant |
One identity holding one source role on one source. |
| Policy rule |
Organisation-wide governance applied after roles allow a request: deny, redact, rate-limit, cap write risk. |
| Catalog |
The recorded structure of each source, from scans. |
| Drift |
What changed in a source’s structure between two scans. |
| Discovery |
Finding content across sources by meaning, from indexed summaries and embeddings. |
| Approval queue |
Where medium and high risk writes wait for a reviewer. |
| Redaction |
Replacing a sensitive value in a response with a marker such as [REDACTED:EMAIL]. |
| Audit log |
One row per governed request, whatever the outcome. |
| Gateway |
The service agents connect to. |
| Admin |
The console and its API. |
| Worker |
The service that runs scans and ingestion. |
