Skip to main content
Watch: I Built an AI App Without Leaving Databricks

Data and BI - official Airbyte server

Airbyte MCP server

Hosted MCP server that connects AI agents to an Airbyte Cloud organization to build, monitor and troubleshoot data pipelines, and (with the context layer enabled) read live data directly from supported sources and destinations. A local version runs against self-managed Airbyte Core via the PyAirbyte package. Airbyte hosts it at https://mcp.airbyte.com/mcp, or you can run airbyte (pypi) locally with OAuth or API key auth.

Last verified 15 tools9 sources

Install

Terminal
claude mcp add --transport http airbyte https://mcp.airbyte.com/mcp
JSON
{
  "mcpServers": {
    "airbyte": {
      "url": "https://mcp.airbyte.com/mcp"
    }
  }
}

VS Code

JSON
{
  "servers": {
    "airbyte": {
      "type": "http",
      "url": "https://mcp.airbyte.com/mcp"
    }
  }
}

Generic

JSON
{
  "mcpServers": {
    "airbyte": {
      "url": "https://mcp.airbyte.com/mcp"
    }
  }
}

Tools

list_cloud_workspaces
List the workspaces you can access.
list_cloud_connectors
List the sources and destinations in a workspace.
deploy_connector_to_cloud
Create a source or destination in a workspace.
check_cloud_connector
Test the configuration and credentials of a source or destination.
list_cloud_connections
List the connections in a workspace.
create_connection_on_cloud
Create a connection between a source and a destination.
set_cloud_connection_selected_streams
Choose which streams a connection syncs.
run_cloud_sync
Start a sync.
get_cloud_sync_status
Get the status of a sync job.
get_cloud_sync_logs
Read the logs from a sync job, to troubleshoot failures.
publish_custom_source_definition
Publish a YAML manifest as a custom connector.
get_agent_skill_docs
Get a connector skill describing entities, actions and parameters for direct reads.
execute_external_api_query
Read live records from a source's API (list and get actions).
execute_external_sql_query
Run a read-only SQL query on a Snowflake or BigQuery destination.
search_airbyte_knowledge_sources
Search Airbyte's documentation, website, API specs, videos and GitHub content (hosted server only).

As documented on October 17, 2026. A running server reports its current list through tools/list.

Notes and caveats

  • Open beta; features and tools may change.
  • Hosted server works only with Airbyte Cloud and Airbyte Flex; sign-in is OAuth 2.0 (SSO supported), or Client-Id/Client-Secret headers from an Airbyte application for headless agents.
  • Direct reads need an org admin to enable the context layer, are read-only, are capped at 100,000 per month per org during beta, and are Cloud-only.
  • Deletes only work on resources whose name contains delete-me or deleteme.
  • For self-managed Airbyte Core, run locally with uvx --from=airbyte@latest airbyte-mcp (stdio, client ID/secret, known issues).
  • Codex is not documented for this server.

FAQ

What does the Airbyte MCP server do?

Hosted MCP server that connects AI agents to an Airbyte Cloud organization to build, monitor and troubleshoot data pipelines, and (with the context layer enabled) read live data directly from supported sources and destinations. A local version runs against self-managed Airbyte Core via the PyAirbyte package. The vendor documents 15 tools, including list_cloud_workspaces, list_cloud_connectors, deploy_connector_to_cloud and check_cloud_connector.

How do I add the Airbyte MCP server to Claude Code?

Run claude mcp add --transport http airbyte https://mcp.airbyte.com/mcp, copied from Airbyte's docs. This page also has snippets for Cursor, VS Code and Generic.

Does the Airbyte MCP server need authentication?

Yes, OAuth or API key. Transport: Streamable HTTP, stdio. Status: Beta, last checked October 17, 2026.

Sources

Last verified against 9 Airbyte pages, then re-checked independently.

Newsletter

New MCP servers, the week they ship

One email per week. New official MCP servers, the coding tools that support them and how developers wire them up.