Creativity · MCP — server

MCP Airbyte Server

The MCP Airbyte Server wraps the Airbyte API (Cloud and OSS) to list sources, destinations, and connections, trigger manual syncs, and read job logs. It's a useful companion for data engineers who want an LLM to summarize which pipelines failed overnight and why — without leaving the chat.

MCP facts

Kind: server
Ecosystem: anthropic-mcp
Language: Python
Transports: stdio, http

Capabilities

Tools: list_connections, trigger_sync, read_job_logs, get_schema, create_connection
Resources: airbyte://connection/{id}, airbyte://job/{id}
Auth: Airbyte API client credentials or personal access token

Install

pipx install mcp-server-airbyte

Configuration

{
  "mcpServers": {
    "airbyte": {
      "command": "mcp-server-airbyte",
      "env": {
        "AIRBYTE_API_URL": "https://api.airbyte.com/v1",
        "AIRBYTE_CLIENT_ID": "${AIRBYTE_CLIENT_ID}",
        "AIRBYTE_CLIENT_SECRET": "${AIRBYTE_CLIENT_SECRET}"
      }
    }
  }
}

Frequently asked questions

Does it work with Airbyte OSS?

Yes — point AIRBYTE_API_URL at your self-hosted instance. Cloud uses client credentials; OSS typically uses basic auth or an API key.

Can it create connectors from scratch?

You can create a connection between an existing source and destination. Defining new custom connectors is best done through Airbyte's Connector Builder UI rather than MCP.

How does it help with on-call?

Pair it with Slack MCP: on a pipeline failure, Claude can pull the latest job logs, diagnose a schema drift, and draft an incident note — all from one conversation.

Sources

Airbyte API documentation — accessed 2026-04-20
Model Context Protocol — accessed 2026-04-20
Airbyte OSS GitHub — accessed 2026-04-20