Airtable's own built-in agents are the first thing to try when the whole job lives inside one base. An outside agent builder earns its place once the agent has to read from, or act in, other apps as well, and an open-source builder suits teams that want to run the automation on their own server.
It's Monday morning and a fresh batch of leads has landed in your base with half the fields empty. Someone has to look up each company, write a one-line summary, tag the status and nudge the right person before the sales call. That's the work people want an AI agent to take off their hands, and it's also exactly where a careless agent can quietly overwrite a column you needed. So before the names, here's what I'd confirm.
What to confirm before an agent writes to your base
Eight checks, most of them answerable from a vendor's integration page, its pricing page or a free account, in the time it takes to drink a coffee.
- Read the integration's list of Airtable actions line by line: you want create and update, not only search and list.
- Test a linked record and a single-select field as well as a plain text column, because a plain text test won't show how the agent handles them.
- Per user, per task, per execution or per credit? Know which unit you're paying in, and what your busiest week would use.
- Find the line on the pricing page that defines one task, run or credit, and ask the vendor if an agent job with several steps counts once or per step.
- Hosted or self-hosted: decide whether the automation runs in the vendor's cloud or on a server you control.
- Connect with an account or token that sees only the bases the agent needs, never your owner login with every workspace.
- Keep a written record of each workflow, because a vendor that closes can delete your account and everything in it.
- Make sure a free plan or trial lets you aim it at a copy of your base, so the first mistakes land on test data.
Built into Airtable, or plugged in from outside
This is the fork that decides most of the rest. Airtable's own AI now lives inside the product: Omni builds apps and automations from a chat, and Field Agents act as researchers, analysts and content writers on the records in a base. If the job is enrichment or summaries within one base, that's the shortest path, because there's no connector to set up and nothing to keep in sync.
Outside tools come in three broad kinds. Hosted agent builders let you describe an agent, give it tools and a knowledge source, and connect it to Airtable alongside your email, CRM and chat apps. Workflow builders give you a visual canvas where AI steps sit next to ordinary triggers and actions, which is better when you want to see every move the agent makes. Portal and app builders sit on top of an existing base and add an AI assistant to the screens your team or clients already use. If you're weighing a move off the base altogether, that's a separate decision, and my guide to Airtable alternatives covers it.
Outside tools reach Airtable through a pre-built integration with a fixed menu of actions. That menu is the real limit. If the action you need isn't on it, the agent can't do it, however clever its model, so that action list is the first page I'd open on any vendor's site.
How Airtable agent tools charge
You'll meet two billing models, and it helps to know which one you're buying into before you build anything. Airtable itself charges per user, with a free plan and paid tiers billed per seat. Its built-in AI rides on that seat model, so the cost question is how many people need a paid seat to use it.
Outside agent platforms usually meter by work instead of people: tasks, workflow executions, runs or credits. Which is cheaper depends on the shape of your job: a small team pushing lots of records through an agent is a very different bill from a large team that runs one job a week. Before you build, ask what a single agent job costs in the vendor's own unit, and whether each record it touches counts separately. Where a free plan exists, it comes with a monthly allowance of tasks or runs, so check that allowance against your real row count rather than assuming it covers a live process.
Open-source automation changes the shape of the bill again: the software is free to self-host, and you pay for the server and your own time, with a hosted paid plan if you'd rather not run it. Exact figures sit on each vendor's pricing page, so price your real monthly volume there, not the headline plan.
A one-hour trial on a copy of your base
Duplicate the base you care about and strip it to a few dozen rows. Pick one job you'd hand to an agent tomorrow, say: read a company name, find a short description, write it to a long-text field and set a status. Connect the tool you're testing to the copy only, using an account that can't see anything else.
Run it on ten rows first. Then check three things. Did every write land in the right field, including linked records? How many tasks, runs or credits did those ten rows use, according to the tool's own usage screen? And when you gave it a row with a missing name, did it stop, guess, or overwrite something? Multiply the usage by your real monthly row count and you've got a far better cost estimate than any plan comparison. A bad run looks like this: summaries written into the notes field instead of the description, a status flipped to done on a row the agent couldn't match, or a linked record replaced rather than added to. Any one of those on ten rows becomes a clean-up job on a thousand. If the agent guessed on the broken row, I wouldn't let it near live data without an approval step.
Nine AI agent tools that work with Airtable
Every tool here can read from or write to an Airtable base. They're grouped by where the agent lives, starting with the platform itself.
Airtable AI (Omni and Field Agents)
Airtable's own AI, built into the product, for building apps by chat and running research and writing agents across the records already in your base.
Zapier Agents
A hosted agent builder for teams that want an AI teammate acting in Airtable and the many other business apps Zapier already connects.
Lindy
An AI assistant you brief in plain English, which can find, create and update Airtable records without you drawing a workflow first.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
Relevance AI
Agent platform offering a ready-made set of Airtable steps for searching, listing and updating records, ready to drop into any agent you build.
StackAI
Small no-code agent builder where an agent can question a base in natural language and write the answers back as records.
n8n
For people who like to see every step, a visual workflow and agent builder you can use hosted or run on your own server.
Activepieces
Open-source automation with an Airtable connector, AI agents and chat-built automations, run in its cloud or self-hosted on your own kit.
Noloco
Portal and internal-app builder that keeps your base in place and adds an AI assistant to the screens on top of it.
Softr
When clients or staff need a front end, this app builder plugs into Airtable as a data source and brings its own built-in agents.
Where I'd start, by the kind of job
If the work stays inside one base, such as summarising, tagging, researching or drafting, start with Airtable's built-in agents. There's no new tool to connect or learn, though anyone using it may need a paid seat, and you'll find out quickly whether AI helps the process at all.
If the agent has to pull from email, a CRM or a form tool and then write back to Airtable, a hosted agent builder or workflow builder is the better fit. Choose the visual workflow kind when you need to audit every step, and the plain-language assistant kind when the people setting it up won't touch a canvas. My round-up of AI agent builders compares those in more depth.
Do you make AI agents or automation tools that connect to Airtable? Tool makers can ask to be considered for a featured place on this list.
If you want the automation running on your own server, look at open-source automation you can host yourself. And if the real goal is giving clients or staff a friendly screen over the base, an app or portal builder with AI included will do more for you than a stand-alone agent.
One last point from this year: an Airtable-connected automation tool, Relay.app, says it shut down in September and deleted all accounts. Whatever you choose, keep a written copy of what each agent does, so you can rebuild it elsewhere in an afternoon if you have to.
Want your tool on this list? Featured spots and reviews here are paid partnerships. Here's how to partner with me.

