Every Airtable base has two halves. The records, which export easily enough, and the structure — table names, field types, select options, which linked-record field points where. Airtable gives you no way to export the second half from the interface.
That matters in three situations: migrating a base somewhere else, handing a base over to someone who has to maintain it, and writing scripts against it.
Where the metadata actually lives
Airtable does expose the structure, just not in the UI. The REST API has a metadata section: one endpoint lists the bases your token can reach, another returns every table in a base along with its fields. Each field comes back with an id, a name, a type, and an options object whose contents depend on the type — the choices of a single select, the precision of a number, the linked table of a linked-record field.
Calling it yourself means creating a personal access token or an OAuth app, scoping it to schema.bases:read, and stitching the responses together. Perfectly doable; rarely worth doing twice.
Getting it in one step
- Sign in to TrilogicPPP with your Airtable account.
- Open the Schema Export tool and pick the base.
- Export. You get the full structure as JSON, with a searchable table list and a size and token-count readout.
- Copy the JSON — or copy the prompt template, which wraps it in the framing an AI assistant needs.
No records are included. That is deliberate: it means the output is safe to paste into a chat window, commit to a repository, or attach to a handover document without leaking customer data.
Schema Export (JSON)
Every table, field, type and option in a base, as JSON. Structure only — no records. With a copy-ready prompt template for AI assistants.
Open the schema exporter →The use case nobody predicted: feeding it to an AI
Ask any assistant to "write me an Airtable script that flags overdue invoices" and it will produce confident, well-structured code referencing Due Date and Status — fields that may not exist in your base, under types it guessed at. The script fails on the first line and you spend the next twenty minutes correcting names.
Pasting the schema first removes the guessing. The model can see that your date field is called Échéance, that Statut is a single select with four specific choices, and that Client is a link to another table rather than a text field. The script it writes then runs.
The token-count readout on the export exists for this reason — a large base's schema can be substantial, and it is useful to know whether it fits in the context window before you paste it. If you would rather skip the copy-paste loop entirely, the Airtable Script Editor already has your schema loaded and runs the generated script against the real base.
Using it for a migration
When a base is moving to Postgres, Notion, or another Airtable workspace, the schema JSON is the specification you rebuild from. A few things to check before you trust it as a complete picture:
- Formula and rollup fields carry their configuration, but the target system will not have the same function set. Treat them as things to reimplement, not port.
- Linked records are symmetric in Airtable. Each side appears as its own field, so a single relationship shows up twice in the JSON.
- Views are not part of the table schema. Filters, sorts and groupings live separately and are not in this export.
- Airtable has no import-schema endpoint, so recreating the base means the create-table API or manual work either way.
Frequently asked
- Does Airtable have a schema export?
- Not in the interface. The metadata is available through the REST API's list-bases and list-tables endpoints, but there is no menu item that hands you a schema file. You either call the API yourself or use a tool that does it for you.
- What is in the schema JSON?
- Every table with its id and name, and for each field its id, name, type and type-specific options — select choices, number precision, date formats, and which table a linked-record field points at.
- Why would I give my schema to an AI assistant?
- Because a model that cannot see your base will invent field names. Pasting the schema first turns a script that fails on the first line into one that references your real fields and types. This is the single most common use of a schema export today.
- Does the schema export include my records?
- No. It describes the structure only, which is what makes it safe to paste into a chat or attach to a handover document.
- Can I recreate a base from the JSON?
- Not automatically — Airtable has no import-schema endpoint. The JSON is the specification you rebuild from, by hand or with a script that calls the create-table API.