Notion Uploader
Pricing
Pay per usage
Upload data into a specified Notion database. It dynamically maps data from any Actor / Dataset to your Notion properties, ensuring your database stays up-to-date with the latest information.
Pricing
Pay per usage
Rating
5.0
(1)
Developer
Filip Cicvárek
Maintained by CommunityActor stats
8
Bookmarked
93
Total users
4
Monthly active users
2.4 days
Issues response
2 months ago
Last modified
Categories
Share
Recommended. An Apify MCP connector authorized to your Notion workspace — create one in Console → Settings → Integrations → MCP connectors. The Actor then acts with your Notion permissions without ever seeing a token. Alternative: leave empty and provide a Notion API key below.
Classic Notion internal-integration secret (starts with 'ntn_' — create at notion.so/profile/integrations). In Notion, connect the integration to the target database/page via the ••• → Connections menu, otherwise it sees nothing. Not needed when a connector is selected above — but adding it anyway enables duplicate lookups (upsert/skip) on Notion plans where the connector cannot run queries.
The Apify dataset whose items get uploaded — pick one from your account or paste a dataset ID. Leave empty when this Actor runs as an integration on another Actor's success: the triggering run's dataset is used automatically.
Upper limit of dataset items to process. 0 = no limit (the whole dataset).
Destination 1 — existing database: paste its Notion URL, a bare ID, or a data-source ID (collection://...). Each dataset item becomes one row. Leave empty when using destination 2 or 3.
For destination 1 only, and only when the database has multiple data sources (a Notion 2025+ feature — the run fails and lists the candidates in that case). Leave empty for normal databases.
Destination 2 — new database: creates it with a schema inferred from your data (column names and types come from the items; leave 'Field mapping' empty to get all fields). Only applies when 'Notion database' above is empty. CAUTION on schedules: every run creates a NEW database unless you paste the created database (URL is in the log and SUMMARY) into 'Notion database' for the next run.
Destination 2 only: the Notion page (URL or ID) under which the new database is created. With an MCP connector you can leave it empty (the database lands as a workspace-level private page); with an API key it is required.
Destination 2 only: title of the created database. Default: 'Apify upload'.
Destination 3 — page: items are rendered as markdown (see 'Page rendering') and written into this page instead of a database. Cannot be combined with 'Notion database'.
Destination 3 only: append to the end of the page, replace the page's content, or create a new child page on every run.
Destination 3 only: how items are rendered — a markdown table of the mapped fields, a bulleted digest, or only the raw text taken from 'Page content field' (set it below in Mapping).
Destination 3 only, when 'Page write mode' is 'Create a new child page per run'. Default: 'Apify upload (N items)'.
Optional plain-English mapping instructions, e.g. 'Put the listing headline into Name, monthly rent into Price, and use the sreality link as URL'. One cheap LLM call resolves it against your dataset fields and the Notion schema — needs the Anthropic API key from the AI features section. Explicit 'Field mapping' below still wins; automatic name-matching fills whatever remains.
Optional explicit mapping of Notion property names to dataset field names, e.g. {"Name": "title", "URL": "url", "Score": "stats.score"}. Dot paths reach nested fields. CAUTION with 'Create a database for me': a non-empty mapping means the new database gets ONLY these columns — leave empty to get all fields.
Destination 1 (existing database) only: create mapped properties that don't exist in it yet. Types are inferred from sample values (text, number, checkbox, URL, date, email, multi-select). New databases get their schema from the data automatically.
Optional dataset field whose text becomes the page body — of each created row (database destinations) or of the target page (page destination with 'Raw content' rendering). Markdown is preserved on the connector path; the API-key path renders headings, bullets and tables.
What to do when an item already exists in the database, matched by the key property below. 'Upsert' updates the existing row, 'Skip existing' leaves it untouched, 'Create' never checks and always adds new rows. Lookups need either a Notion API key or a connector on a Notion Business+ plan — otherwise the run falls back to 'Create' with a warning and 'dedupeDegraded: true' in the SUMMARY.
Required for Upsert / Skip existing: the Notion property that identifies a row, e.g. 'URL'. Works with title, text, number, URL, email, phone and select properties.
Optional: the dataset field holding the identity value. Leave empty to use the field already mapped to the key property.
Optional instruction that rewrites EXISTING field values before upload, e.g. 'append USD to all price values' or 'translate the description to English'. Mechanical changes compile to per-field templates and run locally (one LLM call total); only fields needing judgment get one call per item. The rule plan is saved to the TRANSFORM_PLAN record. Column types adapt automatically in create-database mode; in an existing database, transformed values must still fit the column type. Needs the Anthropic API key.
Optional per-item LLM instruction that GENERATES A NEW VALUE per item, e.g. 'Summarize this listing in one sentence: {{title}}, {{description}}'. {{field}} placeholders are filled from each item and the full item JSON is appended for context. The output lands in the property below. To modify existing values instead, use 'AI value transformation' above. Needs the Anthropic API key.
Only with an enrichment prompt: the Notion text property that receives the AI output. Created automatically when missing.
Needed only for the AI mapping prompt and AI enrichment — get one at console.anthropic.com.
Anthropic model used by the AI features.
Rehearse without writing: resolves the mapping (including the AI prompt), validates against the Notion schema, checks duplicates, and records what WOULD happen (RESOLVED_MAPPING and WOULD_WRITE records, would-create/would-update statuses in the report). No pages are written and no per-item enrichment calls are made. Great first run.
