At a glance: GitHub — 429 stars, TypeScript, MIT license, 114 commits, 15 tools, stdio + HTTP transport, personal access token auth. Community-built by domdomegg (Adam Jones). Not an official Airtable product.

The Airtable MCP server is the most popular community-built MCP integration for Airtable, Howie Liu’s no-code database platform used by 500,000+ organizations. It gives AI agents full CRUD access to bases, tables, records, fields, and comments — everything you need to read, write, and manage Airtable data programmatically through natural language.

Airtable Inc. was founded in 2012 by Howie Liu, Andrew Ofstad, and Emmett Nicholas in San Francisco. The company has raised $1.4B in funding, peaked at an $11.7B valuation (Series F, December 2021), and currently sits around ~$4B on secondary markets. As of early 2026: ~929 employees, ~$478M ARR (up ~27% year-over-year from $375M in 2023). In September 2025, CEO Howie Liu announced a “refounding” of Airtable as an AI-native platform, rebuilding the product architecture with conversational AI as the default interface. Despite this AI push, Airtable does not offer an official MCP server — an Airtable Community thread with 8,700+ views shows strong demand for one.

What It Does

The server exposes 15 MCP tools across four categories:

Database Navigation

Tool What It Does
list_bases Lists all accessible bases with IDs, names, and permission levels
list_tables Retrieves tables from a base with configurable detail levels
describe_table Gets comprehensive info about a table’s structure, fields, and views

Record Operations

Tool What It Does
list_records Query records with optional filtering and record limits
search_records Text-based search across specified or all text fields
get_record Retrieve an individual record by ID
create_record Add new records with specified field values
update_records Modify multiple records simultaneously
delete_records Remove records from tables

Schema Management

Tool What It Does
create_table Create new tables with field definitions
update_table Modify table names or descriptions
create_field Add fields with type specifications
update_field Change field metadata

Collaboration & Attachments

Tool What It Does
create_comment Add record comments with optional threading
list_comments Retrieve comments sorted newest-to-oldest with pagination
upload_attachment Direct file uploads to attachment fields (added v1.11.0)

The server covers Airtable’s core data layer — bases, tables, records, fields, comments, and attachments. It does not support Airtable Automations, Interfaces, Extensions, or Webhooks.

Setup & Configuration

Quick Start (npx)

{
  "mcpServers": {
    "airtable": {
      "command": "npx",
      "args": ["-y", "airtable-mcp-server"],
      "env": {
        "AIRTABLE_API_KEY": "pat..."
      }
    }
  }
}

Other Installation Methods

  • Homebrew: brew install airtable-mcp-server
  • Docker: Container support available
  • DXT: Claude Desktop Extension format
  • Cursor: One-click install from Cursor marketplace
  • Cline: Available in Cline marketplace

HTTP Transport

Set environment variables to run as a stateless HTTP server:

MCP_TRANSPORT=http
PORT=3000

The HTTP transport has no built-in authentication — a reverse proxy with auth is recommended for production use.

Authentication

Uses an Airtable Personal Access Token passed via the AIRTABLE_API_KEY environment variable. You create the token in Airtable’s developer hub with specific scopes:

  • Required: schema.bases:read, data.records:read
  • For writes: data.records:write
  • For comments: data.recordComments:read, data.recordComments:write

The granular scope system means you can create a read-only token for agents that should only query data, or a full-access token for agents that need to create and modify records.

Development History

Date Version What Changed
Dec 12, 2024 Repository created
Dec 22, 2024 v1.1.0–v1.2.0 Added search_records, improved tool guidance
May 11, 2025 v1.5.x View support, sorting, validation fixes
Sep–Oct 2025 v1.7.0–v1.9.0 Rapid iteration, HTTP transport
Nov–Dec 2025 v1.9.5–v1.9.6 Server metadata, stability fixes
Jan 16, 2026 v1.10.0 destructiveHint annotations for dangerous tools
Feb 16, 2026 v1.11.0 upload_attachment tool
Feb 27, 2026 v1.12.0 Typecast support for create/update operations
Mar 7, 2026 v1.13.0 Fixed body size limit for large payloads

15 releases over 15 months with consistent monthly updates. The primary author (domdomegg) has contributed 80 of 114 commits, with Dependabot handling 20 and 7 community contributors adding 1 each. ~15,900 npm downloads per month.

Airtable Pricing

The MCP server is free and open-source. You pay for Airtable’s platform, which determines your API limits:

Tier Price Records/Base API Calls/Month Rate Limit
Free $0 1,000 1,000 5 req/sec/base
Team $20/seat/mo 50,000 100,000 5 req/sec/base
Business $45/seat/mo 125,000 500,000 5 req/sec/base
Enterprise Custom 500,000+ Unlimited 5 req/sec/base

The Free tier is almost unusable for MCP. A single AI conversation — listing tables, searching records, updating a few — can consume dozens of API calls. At 1,000 calls/month, even moderate use would exhaust the allowance in a few sessions. The Team tier (100,000 calls) is the realistic minimum for regular MCP usage.

The hard 5 requests per second per base rate limit applies to all tiers and cannot be increased. There’s also a 50 requests/second limit across all traffic from a given user’s personal access tokens. Exceeding limits returns a 429 status code with a mandatory 30-second backoff.

How It Compares

Feature Airtable MCP (domdomegg) Notion MCP (official) Monday.com MCP (official) Google Sheets MCP (xing5)
Official No (community) Yes Yes No (community)
Stars 429 4,080 383 757
Tools 15 ~20 ~15 ~10
Transport stdio + HTTP stdio stdio + HTTP stdio
License MIT MIT MIT MIT
CRUD Full Full Full Read + write
Schema management Yes Limited Yes No
Comments Yes Yes No No
Attachments Yes No No No
Free tier API limit 1,000/mo Varies Varies Unlimited (Google)
Rate limit 5 req/sec/base 3 req/sec

The biggest gap: Airtable has no official MCP server while both Notion and Monday.com do. Notion’s official server has nearly 10x the stars. For Airtable users, domdomegg’s server is the clear best option, but it lacks the backing and long-term commitment that official servers provide.

Other Airtable MCP Servers

Server Stars Notes
domdomegg/airtable-mcp-server 429 Dominant option, reviewed here
rashidazarang/airtable-ai-agent 4 Python, 33 tools, broader Airtable coverage
onimsha/airtable-mcp-server-oauth 3 Python, adds OAuth 2.1 support
CDataSoftware/airtable-mcp-server-by-cdata 1 Java, read-only, via CData JDBC
glassBead-tc/effect-airtable-mcp 1 TypeScript, Effect-based with type-safe validation

Known Issues

  1. No official backing — This is a community project by an individual developer, not Airtable Inc. If domdomegg stops maintaining it, there’s no corporate fallback. The 8,700+ view community thread shows demand for official support that hasn’t materialized.

  2. Free tier API exhaustion — Airtable’s Free tier allows only 1,000 API calls/month. MCP interactions are API-call-intensive — users report hitting limits quickly even with moderate usage. The Team tier ($20/seat/mo) is realistically required.

  3. Context window consumption — No field filtering on record retrieval. Issue #75 requests support for Airtable’s fields API parameter to limit returned data and conserve LLM context, but this hasn’t been implemented yet.

  4. 5 req/sec hard rate limit — Airtable enforces a 5 requests/second/base limit on all tiers with a mandatory 30-second backoff on 429 responses. An agent doing rapid sequential operations can easily hit this.

  5. No OAuth support — Uses personal access tokens only. A separate fork (onimsha/airtable-mcp-server-oauth) adds OAuth 2.1 support, but the main server requires token management.

  6. HTTP transport lacks auth — Running in HTTP mode exposes the server without built-in authentication. A reverse proxy with auth must be configured separately for any non-local deployment.

  7. Destructive operations — While v1.10.0 added destructiveHint annotations for delete/update tools, the server doesn’t enforce any confirmation flow. An agent can delete records or tables without a safety net beyond the LLM’s judgment.

  8. Limited collaboration features — Comments and attachments are supported, but no access to Airtable Automations, Interfaces, Extensions, Webhooks, or revision history. The server covers data operations, not workflow management.

Bottom Line

The Airtable MCP server is a well-maintained, community-driven integration that covers the essential data operations most users need — listing bases, querying records, creating and updating data, managing schema, and handling comments and attachments. With 429 stars, 15 releases, and consistent monthly updates, it’s the most mature Airtable MCP option available.

The uncomfortable reality is that Airtable — a company positioning itself as AI-native after its September 2025 “refounding” — doesn’t have an official MCP server, while both Notion and Monday.com do. That leaves this community server as the de facto standard, which works well today but carries inherent risk around long-term maintenance.

For practical use: if you’re on the Team tier or above and want AI agents to manage your Airtable data, this server delivers. The 15-tool surface covers CRUD, schema management, and collaboration. If you’re on the Free tier, the 1,000 API calls/month limit makes MCP usage impractical. And if you need automation workflows, webhooks, or advanced Airtable features beyond data operations, you’ll need to supplement with the Airtable REST API directly.

Rating: 4 / 5 — The most complete and actively maintained Airtable MCP server available. Full CRUD with schema management, comments, and attachments. MIT license, multiple installation methods (npx, Homebrew, Docker, DXT), dual transport support. Loses points for lacking official Airtable backing (unlike Notion and Monday.com), no field filtering on record retrieval causing context waste, Free tier API limits that make casual use impractical, and no support for Airtable’s automation and workflow features.

Category: Business & Productivity