Reserved — required by the schema system but unused for cards.
Optional override of the site-wide search provider for this collection. Default: use the active provider from Settings → Search Providers. Text: force text search for this collection (good for SKU-heavy content where exact matches matter).
Controls who can call MCP content tools (query_collection, get_object, search_collection, describe_collection) against this collection. Admin: only callers with a valid API key. Authenticated: callers presenting an OAuth Bearer token with an mcp:* scope (Pro edition). Public: anonymous AI agents can read this collection.
mcp:*
Description shown to AI agents in list_collections and tool descriptions. Falls back to the collection's general description if blank. Write for AI consumption — what is this collection, what kind of objects does it contain, what makes it distinct.
When enabled, objects in this collection are addressable via tcms://{collection}/{id} URIs that AI clients can subscribe to.
tcms://{collection}/{id}
Snake_case identifier AI clients use to call this tool. Becomes the wire-level tool name when registered.
What this tool returns. Shown to AI agents in tools/list — write for AI consumption, not LLM instructions.
Typed caller parameters as JSON. See the docs for the param shape.
Field-name → {value, operator} map as JSON. Filter values may reference {{params.X}} placeholders.
{{params.X}}
REST-style sort spec, e.g. date:desc or priority:asc,date:desc.
date:desc
priority:asc,date:desc
Max results per call. Capped at 50.
Skip the first N matching results.
REST-style include filter — escape hatch for shapes the structured filters object can't express.
REST-style exclude filter — escape hatch.
Output format for rich-text fields.
Saved-query tools exposed to AI agents. Add one per row; each is keyed by its id.