Build a pipeline with AI
Describe in one sentence what you want. Claude composes the pipeline from outsend blocks; you review it before running. (Uses your own Claude key, configured in /settings.)
↓
👉 Click the dot ● on the left of another block
to create the connection · Esc to cancel
Add blocks from the library on the left to build the pipeline.
A pipeline always starts with a 📍 Gmaps Scrape, then chains enrichments in parallel.
The scrape stops once this many businesses are collected. Leave empty for no limit.
Different listings sometimes point to the same number: a shared switchboard, or a network of local listings created to capture calls. Limit it if you want a single contact per number. Listings with no number are never dropped. The same listing never appears twice, whatever this is set to.
Computing EF…
Google Maps requests
GPS points
×
France-equivalent cost
Estimated duration
⚠ Warning: much slower AND many more errors / false positives. Manual or AI-assisted cleanup will be required afterwards.
Only Google Maps scraps less than 7 days old with an available CSV are listed. To reuse an older scrap, run a new scrap or paste the exported CSV.
The Sheet must be shared with public read access ("Anyone with the link can view").
📥 Receives from
filterable input field(s)
No upstream module
Connect a module (Scrap, Import, Enrichment…) to this filter to enable the options.
Connect a module (Scrap, Import, Enrichment…) to this filter to enable the options.
Standard filters
Sampling
Live preview · computing…
Edit a rule to see the preview, or click Recompute.
The previous block hasn't run yet — launch the pipeline to see the preview.
No row matches these rules out of the input rows.
First matched results:
Stats computed on the first 5000 rows (larger CSV).
This module has no settings. It takes the previous block's results as input and produces its own as output.
★ Why this block