You want to point Claude at an Airtable base and say normalize every company name, fill the empty category cells from the description, and tidy the tags. What stops you is everything wired to that base. Views, automations, and interfaces all react the moment a record changes, and the API gives you no diff and no undo. One bad pass ripples through the whole base before you can check it.
Scratch changes where the edit happens. Claude does 99% of the work, reading and rewriting every record as files on your laptop. The last 1%, deciding what actually writes back, stays with you. Nothing reaches the live base until you have seen the change as a diff and approved it.
How it works
- Scratch pulls your base into files. A base or a single table comes down to a folder on your laptop, one file per record, every field laid out for editing.
- Claude edits the records. Open the folder in the Claude desktop app. Try a prompt on a few records in Chat, then let Cowork or Code run it across the whole table. Normalize every company name and fill the empty category cells from the description. Claude edits the files, never the live base.
- You review every diff and publish. Scratch shows each changed field next to the original, word by word. Approve what ships, and Scratch pushes only those records back through the Airtable API.
What people use it for
Most people arrive with a base that has drifted, because tidying it by hand means opening every record one at a time.
- Normalize inconsistent company or contact names across the whole table.
- Fill empty category, status, or tag cells by reading each record's own description.
- Tighten every summary or note down to one line.
- Turn messy free text into the tidy single-select your views expect, without inventing new values.
- Backfill the linked-record references that were left blank.
Pull 50 records to feel the loop, then point Claude at 5,000.
Why not an MCP server?
An Airtable MCP server or extension hands Claude a direct line to your live base. The write button is wired straight in, so one bad pass rewrites every record at once and your views, automations, and interfaces react instantly.
Scratch gives Claude the same full read and write access, but against a local copy. The write-back step is pulled out and handed to you. Claude can change anything; only you can commit it. On a base your team depends on, that is the difference that matters.
What Claude edits in Airtable
- Long-text and rich-text fields
- Single-line text, including titles
- Single-select and multi-select tags
- Date, currency, and number fields
- Attachment captions and linked-record references
Formulas, rollups, lookups, autonumber fields, and created or last-modified timestamps pass through untouched. Validators check your option sets and flag any value that is not in them, right next to the diff. For the full field list, see Scratch for Airtable.
Questions people ask
Is this an MCP server or an Airtable extension?
No. An MCP or extension gives Claude the write button straight to your base. Scratch does not. Claude gets the same access, but writing back is a separate step you approve, one record at a time.
Will Claude break my formulas, views, or automations?
No. Formulas, rollups, lookups, and autonumber fields pass through untouched, automations are never exposed for editing, and validators flag an invented select value right on the diff before you approve it.
Can I undo a change after it writes back?
Yes. Every written record is reversible from Scratch, per row. The original sits next to the rewrite until you decide which one stays.
Why not just bulk-edit with a CSV or a script?
A CSV import and a custom script both write straight to the live base with no diff and no per-record approval, and a find-and-replace only does exactly what you spelled out. Scratch puts an agent on the edit, so it handles the messy cases a rule cannot, and holds every change as a word-level diff you approve before anything ships.
Can it handle a big table at once?
Yes, that is the use case. Pull 50 records to feel the flow, then pull 5,000. Cowork and Code are built for table-scale jobs.
Do I need to be technical?
No. Install Scratch, connect Airtable, point Claude at the folder, and approve the diffs. Validators are optional, for when you run the same job often enough to want guardrails.
See it on your own base
The fastest way to trust it is to watch it run on your data. Book a 30-minute demo on your Airtable base →, or try Scratch free and run the first pass yourself.