agenthub · notion.read
Notion
Read from Notion at a fixed point in the graph — fetch a page, search the workspace, or query a database. A typed data node, not a tool.
Why the node earns its place
A tool hands the model a schema and lets it decide when to call. This is the deterministic counterpart: it reads at a fixed point and emits structured data for downstream nodes to consume.
Credentials never travel on a canvas edge. The node points at a connection by naming it, and the gateway resolves that connection per caller — keyless while you build, the buyer's own Notion at runtime. The same node, keyed per instance.
How it works at run time
- 1Resolve the named connection for whoever is running.
- 2Run the chosen operation:
fetchby id or URL,searchacross the workspace,query_databaseas read-only SQL, orquery_viewto run a view's own filters. - 3Return the envelope: ok, data, error.
What you wire
Inputs
- idstring
- querystring
Outputs
- okboolean
- dataobject
- errorobject
Configuration
| Key | Type | Default & options | What it does |
|---|---|---|---|
| connection * | string | default "notion" | The connection to read through (auth_layer tool_id). Resolved per-caller at run time — keyless in Workbench, the buyer's own Notion in MyCrew. Never a secret. |
| operation * | string | default "fetch" fetch · search · query_database · query_view | fetch: one page/database/data-source by id or URL. search: semantic search across the workspace. query_database: read-only SQL over a data source (Free plan: hourly rate limit). query_view: run a database view's own filters — non-SQL, Free-plan friendly, not rate-limited. Prefer query_view for row lookups on Free. |
| id | string | — | fetch: page / database / data-source id or URL. Accepts a notion.so or *.notion.site URL, a raw UUID, or a collection:// data-source URL. Overridable via the `id` input. |
| include_discussions | boolean | default false | fetch: include comment/discussion markers in the page. |
| query | string | — | search: the semantic query over the workspace. Overridable via the `query` input. |
| teamspace_id | string | — | search: restrict results to one teamspace (optional). |
| page_size | number | default 10 | search: max results to return (1–25). |
| view_url | string | — | query_view: the ?v= URL of a database view. The view's own filters/sorts are applied — use an all-rows view to list everything. Overridable via the `id` input. |
| start_cursor | string | — | query_view: pagination cursor (the next_cursor from a prior page). Leave blank for the first page. |
| data_source_url | string | — | query_database: the collection:// data-source URL. Get it by running a fetch on the database first (from its <data-source url=…> tag). |
| sql | string | — | query_database: read-only SQLite over the data source, e.g. SELECT * FROM "collection://…" LIMIT 10. Requires a Notion Business plan with Notion AI. |
| mcp_tool_name | string | — | Advanced: override the Notion MCP tool name for the chosen operation. Leave blank unless a Notion MCP revision has renamed the tool. |
What usually goes wrong
Watch for this
Prefer query_view over query_database unless you need SQL. SQL requires a Notion Business plan with Notion AI, while a view runs the filters already saved in Notion, is not rate-limited, and works on the free plan.
With no connection the gateway returns a typed not-connected result, surfaced as ok: false — not a crash. Check ok before using data.
Behaviour & provenance