# Quickstart

If you're new to Airtable, the fastest way to get started is to build something from nothing. You don't need to have a pre-defined data model or set up a single table by hand. Describe what you want to track, point your AI tool at whatever raw material you already have, and it builds a real, structured base for you in one sitting.

## Start from scratch: turn a messy idea into a working app for your team

Imagine that you're trying to put together an operations base for your team.

**The sweet spot:** you have a rough idea of what you want to track _and_ a pile of messy source material — a spreadsheet, a folder of PDFs, an email thread, some notes. You can use Airtable's MCP server to turn that messiness into an app for your team.

![A spreadsheet of client project rows with duplicate company names, inconsistent statuses, and mixed date formats](https://static.airtable.com/developers/agent_docs/quickstart_00_messy_spreadsheet@2x.png)

### What you'll need

-   An Airtable account
-   An MCP-compatible AI tool — Claude, ChatGPT, or any connected MCP client.

### Step 1 — Connect Airtable

Add the Airtable connector in your AI tool. If you grant access to a workspace or all content, Airtable will handle everything, from base creation to data ingestion. See our [setup guides](/developers/agents/mcp/getting-started) for connecting to top clients.

![The Airtable authorization screen for an AI tool, with a picker for granting access to everything or to specific workspaces](https://static.airtable.com/developers/agent_docs/quickstart_01_connect_airtable@2x.png)

### Step 2 — Describe what you want, and hand over your mess

Describe what you want to track in plain language and attach or paste whatever raw material you already have. No schema knowledge required.

Don't have anything messy lying around? [Download the sample data here to follow along](https://static.airtable.com/developers/agent_docs/quickstart_sample_client_tracker.csv).

```
Here's what I want to keep track of: the organizations we work with, our contacts at each one, and the projects we're running for them - statuses, owners, and due dates.

And here's what I've got: our team's tracking spreadsheet, attached. Fair warning, it's messy - duplicate companies, inconsistent dates, a few gaps.

Using the Airtable connector, propose a base structure, then once I approve, build it and load my data in.
```

![A chat message describing what to track, with the messy spreadsheet attached as a CSV file](https://static.airtable.com/developers/agent_docs/quickstart_02_describe_what_you_want@2x.png)

### Step 3 — Review the structure before anything is built

Your AI tool proposes the tables, fields, and how they link — for example, three linked tables like **Organizations ↔ Contacts ↔ Projects**, with sensible field types (status, dates, owners, links). Refine it in chat until it's right. Nothing gets created until you approve.

![The AI tool's proposed base structure: Organizations, Contacts, and Projects tables with their fields and links](https://static.airtable.com/developers/agent_docs/quickstart_03_review_the_structure@2x.png)

### Step 4 — Let it build the base and load your data

Once you approve, your AI tool creates the base, tables, fields, and links through the MCP server, then reads your raw inputs and writes them in as clean, de-duplicated, _linked_ records — not a flat dump. It shows you what it's about to write so you can confirm before it commits.

![The AI tool narrating the build while the finished Projects table fills with cleaned, linked records in Airtable](https://static.airtable.com/developers/agent_docs/quickstart_04_build_and_load@2x.png)

## You've got a real base

From here you can:

-   Ask questions across your data — _"Which projects are due in the next two weeks and who's our contact for each?"_
-   Keep feeding it new inputs to add more records over time.
-   Use the server to build Interfaces pages on top of it.

![The AI tool answering a question about upcoming project due dates by querying across the linked tables in the new base](https://static.airtable.com/developers/agent_docs/quickstart_05_ask_your_data@2x.png)

Once you're satisfied, you can go into Airtable and share the base with the rest of your team. No configuration or deployment required. The app is at your fingertips, and theirs. New teammates can query data, create new records, and connect from whatever tools they're using. Airtable is now your team's operational hub.