The New Airbyte Knowledge MCP Server

January 26, 2026

2 min

Summarize with AI:

We're launching a public MCP server that gives AI agents direct access to Airbyte's complete knowledge base. This includes our documentation, API specifications, GitHub issues/PRs, YouTube content, and community discussions.

What's included

The Airbyte Knowledge MCP aggregates information from:

Documentation: Complete technical docs covering setup, connectors, and deployment
OpenAPI specs: Programmatic access to our API structure
GitHub: Issues, discussions, and pull requests across our repositories
YouTube: Tutorial videos and product demos
Website content: Product information and use cases

This gives agents reliable, up-to-date information about how Airbyte works rather than relying on potentially outdated training data.

How to use it

For Claude Desktop or Claude Code, run:

bash

claude mcp add --transport http airbyte-docs

https://airbyte.mcp.kapa.ai

If you're using Cursor or VS Code, you can connect with one click through the Ask AI button in our documentation.

The server works with any MCP-compatible tool including Windsurf and ChatGPT Desktop. The server URL is https://airbyte.mcp.kapa.ai.

When you first connect, you'll authenticate with a Google account (used only for rate limiting, we don't access your personal data). Each user gets 40 requests per hour and 200 per day.

Why we built this

We've been using AI agents internally at Airbyte for development workflows, customer support, and documentation maintenance. We kept running into the same problem: agents would hallucinate outdated information or miss recent changes to our platform.

Building our own MCP server solved this. Now our agents—and yours—can query fresh information directly from our documentation and community resources.

Use cases

The Airbyte Knowledge MCP is particularly useful for:

Building custom connectors: Access code examples and best practices from our GitHub repos
API integrations: Query our OpenAPI specs and implementation guides
Troubleshooting: Search through GitHub issues and discussions for similar problems
Learning Airbyte: Get agents up to speed on our architecture and deployment models

We're already using it internally with our development agents. If you're building agents that interact with Airbyte, give it a try.

Find the full documentation here.

Limitless data movement with free Alpha and Beta connectors

Introducing: our Free Connector Program ->