How to Connect Airtable to an AI Agent, Step by Step

· DoubleX

Quick answer

To connect Airtable to an AI agent, open the Connectors screen, add Airtable, and authorize access to your account. Then pick which actions the agent can use, starting with read-only. In DoubleX Agentes you can run a first test and check the full task history before allowing edits.

Airtable often holds the data a team works from every day: leads, content calendars, inventory, project trackers. Connecting it to an AI agent lets you ask questions about those records in plain language, and, if you choose, have the agent update them. The part that matters most is not the connection itself but deciding what the agent is allowed to touch.

What can an AI agent do with Airtable?#

An AI agent connected to Airtable can read bases, tables, and records, and it can also edit them when you allow it. Reading means searching and summarizing; editing means creating or changing records.

An AI agent is an assistant that uses a language model plus the tools of your systems to do a task for you. Once Airtable is connected, the agent uses that connector's tools instead of you copying data into a chat window.

TaskHow the agent helpsAccess needed
Find recordsSearches a table and lists matchesRead
Summarize a tableDescribes status, owners, and overdue itemsRead
Update a fieldChanges a status or date on a recordEdit
Add a recordCreates a new row from an email or noteEdit

For ideas on what to automate once you are connected, see Airtable with AI agents: what to automate and example requests. This guide stays on the connection and the permissions.

What do I need before connecting?#

You need an account on the platform, an Airtable account that can open the bases you care about, and a clear idea of which base the agent should see first.

A short checklist:

  • An Airtable login with access to the base you want to use.
  • One base to start with, ideally one without sensitive data.
  • A decision on who should be able to use the connection: only you, your team, or the whole company.
  • A first question you can verify yourself, such as "how many records are in this table?"

If you have not created your account yet, creating it is free and needs no card. The Quickstart walks through the first steps, including the diagnostic, a 5 to 10 minute conversation with the product's AI consultant that suggests where to start.

How do I connect Airtable on the Connectors screen?#

Open the Connectors screen, choose Airtable, and sign in to authorize access. After that, the agents you allow can use Airtable's tools.

Step by step in DoubleX Agentes:

  1. Create your account and, if you want, run the diagnostic so the system suggests a team of agents.
  2. Open Connectors and find Airtable in the list.
  3. Start the connection and authorize it with your Airtable account.
  4. Choose whether the connection is personal, for a team, or for the company.
  5. Open the agent that should use Airtable (on the Agents screen) and turn on the Airtable connector for it.
  6. Set which actions that agent may run (next section).

Credentials for connections are stored encrypted with AES-256-GCM. If you want to manage connectors through code instead, the Connectors API reference covers it.

How do I choose which bases and actions the agent can use?#

For each agent, you choose which systems it uses and, inside each system, which actions it can run. The safest start is read-only, with actions that change data turned off.

A practical way to set it up:

  1. Enable only the reading actions first: listing bases, reading tables, and searching records.
  2. Leave actions that create or update records off until the first test works.
  3. Turn on editing only for the agent that needs it, not for every agent.
  4. Keep the agent's instructions specific, for example naming the base and table it should work with.
ApproachWhat the agent can doWhen to use it
Read-onlySearch, list, summarizeFirst test, reporting, audits
Read and editAlso create and update recordsAfter you trust the results

Even with editing on, the agent asks for confirmation before sending, deleting, or changing something in your systems, and it shows the steps it took.

How do I run the first test and check the task history?#

Ask the agent a question you can verify yourself, then open the task and review every step. Each request becomes a task with the request, each tool the agent used, and the result.

Example requests for a first safe test:

  • "List the bases you can access in Airtable."
  • "How many records are in the Leads table of the Sales base?"
  • "Show me the five most recently updated records in the Projects table."
  • "Summarize which records have the status Blocked and who owns them."

Then go to the Tasks screen and open the task. Confirm that the agent used the base you expected and that the numbers match what you see in Airtable. If the answer is wrong, adjust the agent's instructions and try again. Once reading is reliable, turn on one editing action and test it with a single record, approving the confirmation yourself.

Since everything is logged, you can always see who asked, what was done, and when.

Who in the company can see the connection?#

That depends on the scope you choose: a connection can be personal, for a team, or for the whole company. A personal connection is valid only for you.

The platform has owner, admin, and member roles, plus teams and sharing by person, team, or company. A sensible rule: keep the first Airtable connection personal while you test, then move to a team connection once the permissions look right. Each company also has its own separate database, and company content is not used to train models. You can read more in the Introduction or the Connectors API reference.

Frequently asked questions#

Can the agent edit my Airtable records?#

Yes, if you allow it. You choose which actions each agent can run, and you can keep editing off so the agent only reads. Before changing anything, the agent asks for confirmation.

Do I need to code to connect Airtable?#

No. You connect it from the Connectors screen with your Airtable login. Developers who want more control can use the REST API, the CLI, or the MCP server.

Can I limit the agent to one base?#

You control which systems and actions each agent has, and you can name the base in the agent's instructions. Use a narrower Airtable account if you want a stricter boundary on what the connection can reach.

Is my Airtable data used to train AI models?#

No. Company content is not used to train models, and requests go only to AI providers that do not use what they receive for training. A zero-retention option is also available.

Where can I see what the agent did?#

On the Tasks screen. Each task shows the request, each tool the agent used, and the result.