mcp-airtable
Enables AI assistants to perform complete CRUD operations on Airtable bases, tables, records, fields, and comments, with batch processing and file attachment support, plus enterprise-grade reliability features.
README
<div align="center">
<br>
MCP Airtable
<br>
Production-ready MCP server for Airtable integration
<br>
<br>
Built with FastMCP — Works with Claude Desktop & Claude.ai
<br>
<br>
</div>
Overview
A minimal, enterprise-grade MCP server enabling AI assistants to interact with Airtable. Features 21 tools covering complete CRUD operations, batch processing, schema management, and file attachments—built with meticulous precision and strategic clarity.
<br>
Δ Capabilities
<table> <tr> <td width="50%" valign="top">
Core
21 Tools — Full CRUD, batch operations, attachments Streamable HTTP — Claude Desktop 2025+ native support Header Auth — Multi-tenant ready architecture
</td> <td width="50%" valign="top">
Reliability
Circuit Breaker — Cascading failure prevention Auto-Retry — Exponential backoff with jitter Health Checks — K8s liveness & readiness probes
</td> </tr> <tr> <td width="50%" valign="top">
Security
Input Validation — Zod schemas everywhere Injection Prevention — Formula & path attacks blocked Audit Ready — Full security documentation
</td> <td width="50%" valign="top">
Performance
Connection Pooling — Keep-alive via undici Request Deduplication — Shares concurrent results Response Caching — 5-10min TTL for metadata
</td> </tr> </table>
<br>
Quick Start
npm install && npm run build
npm start
Server runs at http://localhost:3000/mcp
<br>
Configuration
<details> <summary><b>Local Development (stdio)</b></summary> <br>
{
"mcpServers": {
"mcp-airtable": {
"command": "node",
"args": ["/path/to/mcp-airtable/dist/index.js", "--stdio"],
"env": {
"AIRTABLE_API_KEY": "patXXXXX.XXXXX..."
}
}
}
}
</details>
<details> <summary><b>Remote HTTP (mcp-remote)</b></summary> <br>
{
"mcpServers": {
"mcp-airtable": {
"command": "npx",
"args": [
"-y", "mcp-remote",
"https://your-server.com/mcp",
"--header", "x-airtable-api-key:patXXXXX.XXXXX...",
"--header", "x-airtable-workspace-id:wspXXXXXXXXXXX"
]
}
}
}
</details>
<details> <summary><b>Claude.ai Web</b></summary> <br>
- Deploy:
npm start - Claude.ai → Settings → Connectors
- Add URL:
https://your-server.com/mcp - Add header:
x-airtable-api-key: patXXXXX...
</details>
<br>
| Platform | Config Path |
|---|---|
| macOS | ~/Library/Application Support/Claude/claude_desktop_config.json |
| Windows | %APPDATA%\Claude\claude_desktop_config.json |
<br>
β Tools
| Category | Tools |
|---|---|
| Bases & Workspaces | list_workspaces° · list_bases · get_base_schema · create_base |
| Tables | list_tables · create_table · update_table |
| Fields | create_field · update_field · upload_attachment |
| Records | get_records · get_record · create_records · update_record · delete_record |
| Batch | upsert_records · delete_records |
| Comments | list_comments · create_comment · update_comment · delete_comment |
| Health | health_check · liveness · readiness |
° Enterprise Scale plan required
<br>
Usage
"List all my Airtable bases"
"Get records from Tasks where Status = 'Active'"
"Create a task with Name='Review PR' and Priority='High'"
"Upload the PDF to the Attachments field on record rec123"
<br>
Authentication
| Priority | API Key | Workspace ID |
|---|---|---|
| 1 | x-airtable-api-key header |
x-airtable-workspace-id header |
| 2 | Authorization: Bearer header |
workspaceId parameter |
| 3 | airtableApiKey parameter |
AIRTABLE_WORKSPACE_ID env |
| 4 | AIRTABLE_API_KEY env |
— |
<br>
Airtable Plan Requirements
<details> <summary><b>API Scopes by Plan</b></summary> <br>
| Scope | Free | Team | Business | Enterprise |
|---|---|---|---|---|
data.records:read |
✅ | ✅ | ✅ | ✅ |
data.records:write |
✅ | ✅ | ✅ | ✅ |
data.recordComments:read |
✅ | ✅ | ✅ | ✅ |
data.recordComments:write |
✅ | ✅ | ✅ | ✅ |
schema.bases:read |
✅ | ✅ | ✅ | ✅ |
schema.bases:write |
✅ | ✅ | ✅ | ✅ |
webhook:manage |
✅ | ✅ | ✅ | ✅ |
user.email:read |
✅ | ✅ | ✅ | ✅ |
workspacesAndBases:read |
❌ | ❌ | ✅ | ✅ |
enterprise.* scopes |
❌ | ❌ | ❌ | ✅ |
</details>
<details> <summary><b>API Rate Limits by Plan</b></summary> <br>
| Plan | Monthly Calls | Rate Limit | Overage Behavior |
|---|---|---|---|
| Free | 1,000/month | 5 req/sec | 30-day grace period (once), then blocked |
| Team | 100,000/month | 5 req/sec | Throttled to 2 req/sec until reset |
| Business | Unlimited | 5 req/sec | — |
| Enterprise | Unlimited | 5 req/sec | — |
</details>
<details> <summary><b>Tool Compatibility</b></summary> <br>
| Tool | Required Scope | Min Plan |
|---|---|---|
list_workspaces |
workspacesAndBases:read |
Enterprise° |
list_bases |
schema.bases:read |
Free |
get_base_schema |
schema.bases:read |
Free |
create_base |
schema.bases:write |
Free |
list_tables |
schema.bases:read |
Free |
create_table |
schema.bases:write |
Free |
update_table |
schema.bases:write |
Free |
create_field |
schema.bases:write |
Free |
update_field |
schema.bases:write |
Free |
get_records |
data.records:read |
Free |
get_record |
data.records:read |
Free |
create_records |
data.records:write |
Free |
update_record |
data.records:write |
Free |
delete_record |
data.records:write |
Free |
upsert_records |
data.records:write |
Free |
delete_records |
data.records:write |
Free |
upload_attachment |
data.records:write |
Free |
list_comments |
data.recordComments:read |
Free |
create_comment |
data.recordComments:write |
Free |
update_comment |
data.recordComments:write |
Free |
delete_comment |
data.recordComments:write |
Free |
° Returns 404 on non-Enterprise plans. Workaround: Get workspace ID from Airtable UI URL (airtable.com/wspXXX/...)
</details>
<br>
Stability & Resilience
<details> <summary><b>Retry with Exponential Backoff</b></summary> <br>
- Auto-retries on HTTP 429, 500, 502, 503, 504
- Handles network errors (ECONNRESET, ETIMEDOUT, ECONNREFUSED)
- Respects
Retry-Afterheaders - Configurable max retries, delays, jitter
</details>
<details> <summary><b>Circuit Breaker Pattern</b></summary> <br>
Prevents cascading failures:
- CLOSED — Normal operation
- OPEN — Fast-fail mode
- HALF_OPEN — Recovery testing
</details>
<details> <summary><b>Request Management</b></summary> <br>
- Timeout: 30s default (AbortController)
- Deduplication: Shares identical concurrent GETs
- Queue: Limits to 5 concurrent requests
- Keep-Alive: Connection pooling via undici
</details>
<details> <summary><b>Health Checks</b></summary> <br>
Kubernetes-ready probes:
health_check— Full status reportliveness— Alive checkreadiness— Traffic ready
</details>
<br>
Deployment
FROM node:20-alpine
WORKDIR /app
COPY package*.json ./
RUN npm ci --production
COPY dist ./dist
EXPOSE 3000
CMD ["node", "dist/index.js"]
docker build -t mcp-airtable .
docker run -p 3000:3000 mcp-airtable
<details> <summary><b>Environment Variables</b></summary> <br>
# Server
PORT=3000
NODE_ENV=production
# Auth (prefer headers in production)
AIRTABLE_API_KEY=
AIRTABLE_WORKSPACE_ID=
# Rate Limiting
RATE_LIMIT_ENABLED=true
RATE_LIMIT_REQUESTS_PER_MINUTE=60
# Caching
CACHE_ENABLED=true
CACHE_TTL_BASES=300
CACHE_TTL_SCHEMA=600
CACHE_TTL_TABLES=300
# Logging
LOG_LEVEL=info
# Sentry (Optional)
SENTRY_DSN=
SENTRY_DEBUG=false
SENTRY_ENVIRONMENT=production
</details>
<br>
Architecture
src/
├── index.ts # Entry point
├── server.ts # FastMCP init
├── tools/ # 21 tools
│ ├── bases.ts
│ ├── tables.ts
│ ├── fields.ts
│ ├── records.ts
│ ├── batch.ts
│ └── comments.ts
└── lib/ # Core
├── airtable/ # API client
├── retry.ts # Backoff
├── circuit-breaker.ts # Failure prevention
├── health.ts # K8s probes
├── deduplication.ts # Request dedup
└── ...
~2,500 lines · 272 unit tests · 25 e2e tests
<br>
References
| MCP Specification | modelcontextprotocol.io |
| FastMCP | github.com/punkpeye/fastmcp |
| mcp-remote | npmjs.com/package/mcp-remote |
| Airtable API | airtable.com/developers |
<br>
<div align="center">
<br>
Security · Examples · Changelog
<br>
MIT License
<br>
DELTΔ & βETΑ
From Change to What's Next
<br>
</div>
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。