airbyte-mcp

airbyte-mcp

Enables interaction with an Airbyte instance through natural language, supporting workspaces, sources, destinations, connections, jobs, logs, tags, streams, and connector definitions.

Category
访问服务器

README

Airbyte MCP Server (airbyte-mcp)

MCP server for the Airbyte Public API. Built with the official MCP Python SDK (FastMCP).

Lets any MCP-compatible client (Cursor, Claude Desktop, Claude Code, MCP Inspector, etc.) interact with your Airbyte instance through natural language.

Features

  • 30 tools covering workspaces, sources, destinations, connections, jobs, job logs, tags, streams, and connector definitions
  • Read and write operations for core resources (create, update, delete)
  • Job diagnostics via the internal Configuration API (self-managed): detailed failure reasons, per-stream stats, and structured logs
  • Automatic token exchange with in-memory caching and transparent 401 retry
  • Markdown and JSON response formats on summary tools; JSON-only for log tools
  • Pagination support (limit/offset) on all list tools
  • Two transport modes: stdio (local) and streamable HTTP (remote)
  • Works with self-managed Airbyte (abctl) and Airbyte Cloud

Available Tools

Tool Description
Health
airbyte_health_check Ping the Airbyte API
Workspaces
airbyte_list_workspaces List workspaces with pagination
airbyte_get_workspace Get workspace details by ID
Sources
airbyte_list_sources List source connectors (filter by workspace)
airbyte_get_source Get source details by ID
airbyte_create_source Create a new source connector
airbyte_update_source Update an existing source
Destinations
airbyte_list_destinations List destination connectors (filter by workspace)
airbyte_get_destination Get destination details by ID
airbyte_create_destination Create a new destination connector
airbyte_update_destination Update an existing destination
Connections
airbyte_list_connections List connections / pipelines (filter by workspace)
airbyte_get_connection Get connection details including stream config
airbyte_create_connection Create a new connection (pipeline)
airbyte_update_connection Update an existing connection
Jobs
airbyte_list_jobs List jobs (filter by connection, type, status, dates)
airbyte_get_job Get job details (status, duration, bytes/rows synced)
airbyte_trigger_sync Trigger a sync or reset job
airbyte_cancel_job Cancel a running job
Job Logs (Internal API)
airbyte_get_job_details Per-attempt stats, failure reasons, and stacktraces
airbyte_get_job_logs Structured log entries for all attempts
airbyte_get_attempt_logs Structured log entries for a specific attempt
Streams
airbyte_get_stream_properties Get stream properties for a source/destination pair
Tags
airbyte_list_tags List tags
airbyte_create_tag Create a tag
airbyte_update_tag Update a tag
airbyte_delete_tag Delete a tag
Connector Definitions
airbyte_list_source_definitions List source connector definitions
airbyte_get_source_definition Get a source connector definition
airbyte_list_destination_definitions List destination connector definitions
airbyte_get_destination_definition Get a destination connector definition

See docs/endpoints.md for the full Airbyte API endpoint checklist.

Prerequisites

Quickstart

1. Clone and install

git clone https://github.com/trustxai/airbyte-mcp.git
cd airbyte-mcp
uv sync

2. Configure credentials

cp .env.example .env

For self-managed (abctl), retrieve credentials:

abctl local credentials

Edit .env with your client-id and client-secret. See docs/authentication.md for details.

3. Run the server

stdio (Cursor / Claude Desktop / Docker)

uv run airbyte-mcp

Streamable HTTP (remote / MCP Inspector)

uv run airbyte-mcp-http
# Listening on http://127.0.0.1:8080/mcp

Client Configuration

Every MCP client (Cursor, Claude Desktop, etc.) can run the server in one of two ways:

  • uv — quickest to set up, but requires uv installed locally.
  • Docker — no Python / uv required on the host; everything runs in a container. Build the image once and every client config reuses it.

Host networking note: if Airbyte is running on your host machine (e.g. via abctl), inside the Docker container localhost does not point to your host. Use http://host.docker.internal:8000/api/public/v1 on macOS/Windows, or add --network=host to the docker run args on Linux.

Build the Docker image (one-time)

docker build -t airbyte-mcp:latest .

Cursor

Add to .cursor/mcp.json (project-level) or ~/.cursor/mcp.json (global):

Option A — uv

{
  "mcpServers": {
    "airbyte": {
      "command": "uv",
      "args": ["--directory", "/path/to/airbyte-mcp", "run", "airbyte-mcp"],
      "env": {
        "AIRBYTE_API_URL": "http://localhost:8000/api/public/v1",
        "AIRBYTE_CLIENT_ID": "<your-client-id>",
        "AIRBYTE_CLIENT_SECRET": "<your-client-secret>"
      }
    }
  }
}

Option B — Docker

{
  "mcpServers": {
    "airbyte": {
      "command": "docker",
      "args": [
        "run", "--rm", "-i",
        "--name", "airbyte-mcp",
        "-e", "AIRBYTE_API_URL",
        "-e", "AIRBYTE_CLIENT_ID",
        "-e", "AIRBYTE_CLIENT_SECRET",
        "airbyte-mcp:latest"
      ],
      "env": {
        "AIRBYTE_API_URL": "http://host.docker.internal:8000/api/public/v1",
        "AIRBYTE_CLIENT_ID": "<your-client-id>",
        "AIRBYTE_CLIENT_SECRET": "<your-client-secret>"
      }
    }
  }
}

Claude Desktop

Add to claude_desktop_config.json:

Option A — uv

{
  "mcpServers": {
    "airbyte": {
      "command": "uv",
      "args": ["--directory", "/path/to/airbyte-mcp", "run", "airbyte-mcp"],
      "env": {
        "AIRBYTE_API_URL": "http://localhost:8000/api/public/v1",
        "AIRBYTE_CLIENT_ID": "<your-client-id>",
        "AIRBYTE_CLIENT_SECRET": "<your-client-secret>"
      }
    }
  }
}

Option B — Docker

{
  "mcpServers": {
    "airbyte": {
      "command": "docker",
      "args": [
        "run", "--rm", "-i",
        "--name", "airbyte-mcp",
        "-e", "AIRBYTE_API_URL",
        "-e", "AIRBYTE_CLIENT_ID",
        "-e", "AIRBYTE_CLIENT_SECRET",
        "airbyte-mcp:latest"
      ],
      "env": {
        "AIRBYTE_API_URL": "http://host.docker.internal:8000/api/public/v1",
        "AIRBYTE_CLIENT_ID": "<your-client-id>",
        "AIRBYTE_CLIENT_SECRET": "<your-client-secret>"
      }
    }
  }
}

Claude Code (HTTP)

Option A — uv

uv run airbyte-mcp-http &
claude mcp add --transport http airbyte http://127.0.0.1:8080/mcp

Option B — Docker

docker build -f Dockerfile.http -t airbyte-mcp-http:latest .
docker run -d --rm --name airbyte-mcp-http \
  -p 8080:8080 \
  -e AIRBYTE_API_URL=http://host.docker.internal:8000/api/public/v1 \
  -e AIRBYTE_CLIENT_ID=<your-client-id> \
  -e AIRBYTE_CLIENT_SECRET=<your-client-secret> \
  airbyte-mcp-http:latest
claude mcp add --transport http airbyte http://127.0.0.1:8080/mcp

MCP Inspector

# Start the server in HTTP mode (uv or docker — either works)
uv run airbyte-mcp-http
#   or: docker run --rm -p 8080:8080 --env-file .env airbyte-mcp-http:latest

# In another terminal
npx @modelcontextprotocol/inspector
# Then connect to http://127.0.0.1:8080/mcp in the inspector UI

Running Manually (without a client)

If you just want to exercise the server from the CLI:

# stdio (uv)
uv run airbyte-mcp

# stdio (Docker)
docker run --rm -i --env-file .env airbyte-mcp:latest

# HTTP (uv)
uv run airbyte-mcp-http

# HTTP (Docker)
docker run --rm -p 8080:8080 --env-file .env airbyte-mcp-http:latest

Environment Variables

Variable Required Default Description
AIRBYTE_API_URL No http://localhost:8000/api/public/v1 Airbyte API base URL
AIRBYTE_CLIENT_ID Yes* Application client ID
AIRBYTE_CLIENT_SECRET Yes* Application client secret
AIRBYTE_ACCESS_TOKEN No Pre-fetched token (skips exchange)
HTTP_HOST No 127.0.0.1 HTTP transport bind address
HTTP_PORT No 8080 HTTP transport port

*Not required if AIRBYTE_ACCESS_TOKEN is provided.

Documentation

Contributing

Contributions are welcome! See docs/CONTRIBUTING.md to get started.

License

Apache-2.0 — see LICENSE for details.

推荐服务器

Baidu Map

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。

官方
精选
JavaScript
Playwright MCP Server

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。

官方
精选
TypeScript
Magic Component Platform (MCP)

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。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

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。

官方
精选
本地
TypeScript
VeyraX

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。

官方
精选
本地
graphlit-mcp-server

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。

官方
精选
TypeScript
Kagi MCP Server

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。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

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。

官方
精选