Using the Airtable MCP server

  • Updated on Apr 30, 2026
  • Published on Feb 11, 2026
  • 11 minute(s) read
Prev Next

Plan availability

All plans

Permissions

Your MCP permissions mirror the same permission level(s) that you have have been granted in Airtable.

  • Owner, Creator, or Editor - Can set up OAuth integrations and manage MCP access. Can read and update records they have permission to access.

    • Additionally, workspace owners or creators can create new bases via MCP. Note that base-level creator permissions will not be able to create a new base.

  • Commenter or Read-only - Can read data via MCP for bases they have access

Platform(s)

Works with AI assistants that support Model Context Protocol (Claude, ChatGPT, and other MCP-compatible tools)

Airtable’s MCP server lets you connect AI tools directly to your Airtable bases. This integration allows you to ask questions about your data, create and update records, and analyze information—all through natural conversation with AI tools like Claude, ChatGPT, Cursor, and more.

What is Model Context Protocol (MCP)?

Model Context Protocol (MCP) is an open standard that allows AI tools to securely connect to external data sources. With MCP, AI tools can query your Airtable data, perform actions on your behalf, and present structured information—all while respecting your existing permissions and security settings.

What you can do with MCP?

The Airtable MCP Server enables several types of interactions between your AI tools and your data:

  • Search and analyze data - Ask your AI tool questions about your Airtable data:

    • "Show me all customers from our CRM base who haven't been contacted in the last 30 days"

    • "What are the top 5 most common feature requests in our feedback tracker?"

    • "Find all projects with status 'At Risk' and summarize the blockers"

  • Create records - Add new information to your bases through conversation:

    • "Add these 5 new project tasks to my Project Tracker base"

    • "Create a new contact for Jane Smith at Acme Corp with email [email protected]"

  • Update existing records - Make changes to your data without opening Airtable:

    • "Update the status of task #123 to 'Complete'"

    • "Change the due date for the Q1 Planning project to March 31st"

  • Create bases - Set up new bases directly from your AI tool:

    - "Create a new base called 'Q3 Campaign Tracker' in my Marketing workspace"

    - "Set up a project tracker base with tables for Tasks, Milestones, and Team Members"

  • Access interface data - Query data from Airtable Interfaces you have access to, even if you only have interface-level permissions:

    • "What are the open tasks on my project tracker interface?"

    • "Show me the records from the Q1 Pipeline page"

MCP prerequisites

Before setting up the Airtable MCP Server, you'll need:

  • An MCP-enabled Host (Claude, ChatGPT, Cursor, etc)

  • Data stored in Airtable bases

  • Permission to use third-party integrations in your Airtable organization (if you don’t have permission, an Airtable admin at your organization may need to allow-list a new integration)

Tips for optimal MCP usage

Get the most out of your MCP integration with these tips:

  • Be specific: When asking about data, mention the base name or describe what you're looking for clearly

  • Use natural language: You don't need to know field names or table IDs—describe what you want in plain English. Sometimes you may want to describe the base in more detail before asking for data from a table.

  • Check permissions: If you can't access certain data via MCP, verify that you have the appropriate permissions in Airtable.

  • Think about what data you’re giving your AI tool: Your tool will be able to read and make edits to your Airtable data on your behalf. Make sure you trust the tools you’re using or any other data sources you connect to.

Setting up a quick MCP connection

The following tools have a pre-built connector to Airtable and can be connected in a single step:

  • ChatGPT - Open AI’s chat assistant

    • Learn more here

    • Client ID for allowlisting: 7a713e1a-3d99-4fdf-b59a-311bdf94ba97

  • Claude - Anthropic's chat assistant

    • Learn more here

    • Client ID for allowlisting: 266cb1c0-b4ae-43a1-b7d7-c2a563667d95

  • Claude Code - Anthropic's agentic coding tool

    • Run this command in your terminal: claude mcp add --transport http airtable https://mcp.airtable.com/mcp

    • Inside Claude Code, run /mcp to authenticate with your Airtable account.

    • This is the recommended setup path. Users can also connect via personal access token — see the PAT section below.

MCP setup steps (manual connection)

These are the manual steps to follow if your AI tool doesn’t have a pre-built connector. There are two approaches to consider.

Use this approach if your AI tool asks you to manually enter a client ID and secret during setup.

  1. Follow the Airtable OAuth application setup flow to register an OAuth client.

    1. Your tool should supply the redirect URI required to create the app

    2. Create the app with the following scopes:

      • data.records:read

      • data.records:write

      • schema.bases:read    

      • schema.bases:write

      • data.recordComments:read

      • data.recordComments:write

      • workspacesAndBases:read

      • webhook:manage

  2. Copy your client ID (and, optionally, client secret).

  3. In your AI assistant's MCP configuration, enter:

  4. Complete the OAuth authorization flow.

Note

OAuth clients created this way can be shared across multiple users in your organization if you add a support email, terms of service and privacy policies.

Personal Access Token registration

Alternatively, you may also authenticate with a personal access token if your AI tool supports custom request headers or authentication tokens.      

  1. Follow the Airtable personal access token (PAT) setup flow to register a PAT.

  2. Create the app with the following scopes:

    1. Data.records:read

    2. Data.records:write

    3. schema.bases:read

    4. schema.bases:write

    5. data.recordComments:read

    6. data.recordComments:write

    7. workspacesAndBases:read

    8. webhook:manage

  3. Copy your PAT.

  4. In your AI assistant's MCP configuration, enter:

    1. Server URL: https://mcp.airtable.com/mcp

    2. Custom “Authorization” header: “Bearer [your personal access token]”

    3. For Claude Code using a personal access token (instead of OAuth): claude mcp add --transport http airtable https://mcp.airtable.com/mcp --header "Authorization: Bearer [your personal access token]"

      1. Note that we recommend the OAuth path of registering the MCP, however, users who prefer token-based auth or have environments where browser-based OAuth isn't practical can utilize this method.

Managing your MCP integration

After setup, you can control which bases your AI assistant can access:

  1. Navigate to your third-party integrations with this link, or click on your user profile in Airtable and click Integrations and then select Third-party Integrations.

  2. Select your MCP integration.

  3. Add or remove bases/apps as needed.

    1. You can also add apps where you have interface-only access.

Available MCP capabilities

Currently, the MCP server provides these tools:

Tool name

Description

ping

Ping the MCP server to check if it is running

create_table

Creates a new table in an Airtable base. To get the baseId, use the list_bases or search_bases tools first.

create_field

Creates a new field in an existing Airtable table. To get baseId and tableId, use the search_bases and list_tables_for_base tools first.

create_base

Creates a new base in a specified workspace. To get workspaceId, use the list_workspaces tool first. You can optionally define tables and fields at creation time; if omitted, a default table is created.

create_records_for_table

Creates new records in an Airtable table. To get baseId and tableId, use the search_bases and list_tables_for_base tools first.

update_records_for_table

Updates records in an Airtable table. The fields you specify will be updated, and all other fields will be left unchanged. To get baseId and tableId, consider using the search_bases and list_tables_for_base tools first.

update_table

Updates an existing table's name and/or description in an Airtable base. To get baseId and tableId, use the search_bases and list_tables_for_base tools first.

update_field

Updates the name and/or description of a field in an existing Airtable table. To get baseId, tableId, and fieldId, use the get_table_schema tool first.

list_bases

Lists all bases (applications) that you have access to in your Airtable account. Use this to get the baseId of the base you want to use. Favorited and recently viewed bases are generally more relevant.

list_workspaces

Lists all workspaces you have access to in your Airtable account, along with your permission level in each. Use this to get the workspaceId needed for tools like create_base.

list_records_for_table

Lists records queried from an Airtable table. Do not assume baseId and tableId. Obtain these from search_bases → list_tables_for_base. Do not attempt to pass filterByFormula. Look carefully at the filters parameter.

Pre-requisite: If filtering on select/multiSelect fields, you must call get_table_schema first to get the choice IDs. Aim to provide at least 6 relevant fields via the fieldIds parameter.

list_tables_for_base

Gets the summary of a specific base (application). This includes the schemas of all tables in the base, including field name and type.

list_pages_for_base

Lists all interfaces and their pages for a base, including page IDs, names, the tables each page displays, and the visible fields on each table. Use this to discover interface pages, especially for bases where you have interface-only access.

list_records_for_page

Lists records from an interface page. Supports simple pages and hierarchy pages (e.g. projects → tasks). Use this for bases where you have interface-only access, or when querying data through a specific interface page.

display_records_for_table

Note: This tool is not available by default and is only enabled for clients that support interactive apps. If you are a developer who wants to use this tool, please fill out this form.

Displays an interactive widget showing record data queried from an Airtable table. Do not assume baseId and tableId. Obtain these from search_bases → list_tables_for_base. Do not attempt to pass filterByFormula. Look carefully at the filters parameter.

get_record_for_page

Gets a single record's details from an interface page element/visualization.

get_table_schema

Gets the detailed schema information for specified tables and fields in a base. This returns the field ID, type, and config for the specified fields of the specified tables.

search_records

Searches for records in a table using free-text search.

search_bases

Searches for bases (applications) by name. This is useful when you need to find a specific base quickly by a partial name-based match. Returns bases sorted by their relevance score, as well as a recommended base ID and a hint on whether we need to ask the user to explicitly select the base they want to use.

Adding agent skills to use with Airtable’s MCP server

Agent skills are basically downloadable markdown files that users can use with AI clients to provide them with instructions or extra information to help them work more effectively with Airtable’s MCP server. Currently, there are two skills, but we plan on adding more in the future:

  • airtable-overview - This skill explains Airtable's data model so your agent understands how bases, tables, and fields relate.

  • airtable-filters - This skill improves record retrieval, making queries more accurate and reliable.

To install skills:

  1. Visit this official Airtable agent skill GitHub repository and download one or more skills.

  2. Next, you’ll need to manually install the skill(s) within your AI platform. Depending on your AI client, this will involve slightly different steps, so please refer to your AI platform’s unique instructions for installing skills.

    1. Some AI platforms require skills to be uploaded one at a time. This means users are unable to directly paste the parent folder .zip file to install all skills at the same time. For these clients, users will need to manually upload each individual SKILL.md file present in the .zip folder.

    2. You may need to ask your organization’s AI admins to allow skills before you upload skills to your AI platform.

  3. Once one or more skills are uploaded, your AI platform/model is usually in charge of when it gets activated and used. From your perspective, you don’t have to do anything else; your Airtable AI workflows will be able to invoke these tools to produce better outputs. However, some AI platforms also allow users to manually “invoke” skills. Check your AI platform’s documentation for more information about manual skill usage.

Limitations and dependencies

Keep these limitations in mind while using the MCP Server:

  • Rate limits: MCP server calls are subject to standard Airtable API rate limits.

  • Batch size: Creating records is limited to 10 records per request.

  • Change policy: Tool names, behaviors, and capabilities may change

  • Permission requirements: The MCP server respects your existing Airtable permissions—you can only access data you normally have permission to view or modify.

  • Third-party integrations: If you are an enterprise user and your admin has enabled restrictions on third-party integrations, they'll need to allowlist the MCP integration.

FAQs

Can I use MCP with any base I have access to?

If you have base-level access, you can use MCP to access the data stored within that base. The MCP server respects your existing Airtable permissions. If you have read-only or comment-only access to a base, you'll be able to read that data via MCP, but you won't be able to create or update records.

What AI assistants work with Airtable's MCP Server?

Any AI assistant that supports the Model Context Protocol can connect to Airtable's MCP Server. This includes Claude, ChatGPT, and other MCP-compatible tools. Check your AI assistant's documentation to confirm MCP support.

How do I revoke access if I no longer want my AI assistant to access my bases?

Navigate to your user profile > Integrations > Third-party Integrations, find the MCP integration, and either remove specific bases or delete the integration entirely.

Does using MCP count against my Airtable limits?

MCP uses our public API under the hood and will count against any existing API rate limits. Any records created using MCP will count against existing base limits.

Can I use MCP to access Airtable Interfaces?

Yes. If you have interface-only access to a base, you can use MCP to discover your interface pages (list_pages_for_base), list records shown on a page (list_records_for_page), and drill into record details (get_record_for_page). The data returned respects the same permissions as your interface — you'll only see data that your interface pages make available to you.

What happens if I try to update a field I don't have permission to modify?

The MCP server will return an error, just as the Airtable API would. Your AI assistant should communicate this limitation to you.

Are there additional costs for using MCP?

No. The MCP integration is currently included with your current Airtable plan. There are no additional Airtable charges for MCP usage, but existing API rate limits remain.

How can admins enable a blocked MCP integration?

For Airtable admins to enable a blocked integration:

  1. Navigate to Admin Panel > Settings > Integrations & development > Third party integration allowlist.

  2. Click Allow integration.

  3. Enter the client ID provided by your user.

  4. Click Save.

Can I create new bases using MCP?

Yes. You can use the create_base tool to create a new base in any workspace where you have Creator-level permissions. You can optionally define tables and fields during creation. Use list_workspaces first to find the workspace ID where you want to create the base.

Can the Airtable MCP connect to development apps (part of the managed apps and components feature) and update those apps in the same way as regular Airtable bases?

No. Calls to a development base will return a 403 permissions error. However, API/MCP calls can be made to bases where a managed app or component has been installed (i.e., the child/production bases). So the restriction is specifically on the development base itself, not on installed instances.

How do I connect Claude Code to Airtable?

The recommended approach is to run claude mcp add --transport http airtable https://mcp.airtable.com/mcp in your terminal, then run /mcp inside Claude Code to authenticate with your Airtable account through your browser. Alternatively, you can connect using a personal access token if you prefer — see the Personal Access Token section above for instructions.