Skip to main content
This page maps the REST surface resource by resource: how to list, get, create, update, and terminate each one, and — because most requests reference other resources — exactly where each referenced ID comes from. Cells marked not exposed are deliberate statements, not omissions; the gaps table collects them with the supported alternative. Two published surfaces serve these resources: The older polymorphic /api/entities route is a compatibility adapter. It is documented separately for existing clients and is not part of the v1 OpenAPI contract.

Naming crosswalk

The same record can have a concept name and a wire name. This table is the authority when the names differ: Where the wire uses snake_case (task_id, receipt_id) it is a request or portable-document field; where it uses camelCase (taskId, receiptId) it is a /api/v1 response field. Both refer to the same records.

Canonical v1 resources

These are the resources a new integration should build against. Every path in this table is in the published OpenAPI document.

Compatibility-only entity reads

The polymorphic /api/entities route remains available for older MCP and workspace integrations. It is not a v1 resource namespace and is not included in the table above. Use it only when a legacy client requires a record that has no dedicated v1 read. MCP column, condensed: orgx_write and orgx_search use the compatibility adapter for records that do not yet have a dedicated MCP operation; orgx_act, orgx_decide, and review_artifact carry human-authority flows; orgx_spawn creates runs. Operating processes, handoffs, discovery runs, episodes, the event stream, and projections are REST-first. The full mapping is on the MCP ↔ REST parity page.

Where every ID comes from

Not addressable with an API key

Honest inventory of what a developer credential cannot do today, with the supported route: If an operation is absent from both this map and the OpenAPI document, treat it as unsupported.