MCP connectorsAirtable

Airtable MCP: connect Airtable to Claude and ChatGPT

Ask Claude or ChatGPT which deals in your pipeline base were won this month, which suppliers have no contact person, or what the team commented on a record. The answer comes straight from Airtable. No building a new view just to answer one question.

With Deployed OS everyone connects their own Airtable login and picks the bases to share. The AI works in exactly those, with that person’s rights, and only adds a record when someone asks and confirms the values.

4 read actions · 1 write action · Works in Claude, ChatGPT and Copilot

In practice

What your team can do with Airtable in Claude and ChatGPT

Sales

The pipeline without a new view

Ask for the deals in Negotiation, sorted by close date, with amount and owner. The AI writes the Airtable formula, so you don’t have to.

Operations

Trackers that answer back

Suppliers, equipment, orders: whatever your team keeps in Airtable. Ask what’s due, what’s missing a field or who owns it, and get the records back with their comments.

Marketing

The content calendar on demand

What publishes next week, which posts are still in Draft and what the comments on each one say. One answer before the Monday meeting.

Any team

Add a record from the chat

A new lead from an email, a bug from a call: ask the AI to add it to the right table. It shows you the values first and uses the select options you already have.

Examples

Ask it like this

“In the Sales pipeline base, which deals are Won with a close date this month? Total the Amount field.”
“Which suppliers in the Vendors table have no contact person filled in?”
“What fields does the Content calendar table have, and which options does Status have?”
“Show the records in the Next two weeks view of the Launch plan table, sorted by date.”
“What did people comment on the Nordbyte record in our CRM base?”
“Add a lead to the Leads table: Name Sara Lind, Company Nordbyte, Source Webinar, Status New.”

The connector

What the Airtable MCP connector does today

Readsairtable_list_bases
List the Airtable bases you shared when you connected, with your permission level in each.
Readsairtable_list_tables
List a base’s tables with their fields, field types, select options and views.
Readsairtable_list_records
List a table’s records, up to 100 a page: in a view, filtered by an Airtable formula, with chosen fields and a sort.
Readsairtable_get_record
Read one record: its fields, its latest comments and its link when there is one.
Writesairtable_create_record
Create one record with values by field name, where a select field takes an existing option. Nothing else changes and nobody is notified, but the base’s automations for new records still run.

Need it to do more? We extend connectors for the companies that use them.

Security

Who sees what

Read, plus new records: The connector reads bases, tables, fields, records and record comments. It can create one record at a time. It cannot update or delete records, or change fields, tables or views.
Only the bases people share: Everyone connects their own Airtable login and picks which bases to share. The AI works in exactly those, with that person’s permission in each. A base left out stays out.
You choose who gets it: Admins decide which groups at the company get the Airtable connector, and which of them may create records. Everyone else never sees it.
Every call logged: Who asked, which connector, when. Deployed OS shows reads and writes per connector and per person, and the log can be exported. When someone leaves, access is revoked centrally.

Setup

Set up the Airtable connector in three steps

An admin registers an OAuth integration at airtable.com/create/oauth, with the redirect URL and scopes from Deployed OS, and generates a client secret. Your company owns the integration, not us.

The admin adds its Client ID and client secret to Deployed OS and chooses which groups get Airtable.

Each person connects with their own Airtable login and picks the bases to share. From then on they just ask.

Compared

Airtable’s own MCP server or Deployed OS?

Airtable runs its own hosted MCP server, and a person connects it with their own Airtable login, choosing what to share. It is a good fit for one person, or a developer working in Airtable from their own AI tools.

Deployed OS is built for the whole company. One company connection for Airtable and your other tools, each person’s own login, admins choosing which groups get what, and a log of every call. The same connection works in Claude, ChatGPT and Copilot, next to your company’s skills and guardrails.

Their documentation: Airtable’s MCP server

Questions

Airtable MCP: common questions

Yes. Airtable hosts its own MCP server, which a person connects to their own AI tool. Deployed OS gives the whole company one connection, with each person’s own login, access per group and a log of every call.

Yes. Deployed OS is your company’s own MCP, so the same Airtable connection works in ChatGPT, Claude and Copilot. People keep the assistant they prefer.

Only by adding records, one at a time, with the values you confirmed and the select options you already have. Your company’s guardrails apply. Nothing is updated or deleted and nobody is notified, but the base’s own automations for new records still run. If you need more, for example updating records, we can extend it for you.

No. Ask in plain words and the AI writes the filter for you, for example {Stage} = "Won", using the field names in your table.

Because it wasn’t picked when that person connected. Airtable asks which bases to share, so reconnect Airtable in Deployed OS and add it.

It depends on the size of your team. Get in touch and we will walk you through it.

More connectorsAll connectors

Let’s talk

See it on your own tools.

Tell us which systems your company runs on. We’ll show you what your team can ask, and what IT gets to control.

You’ll hear from Poyan Karimi

Co-founder · Deployed AI