A few weeks back we launched fast-pyairbyte, a remote MCP server which allows developers to generate an entire date pipeline with a single prompt. The MCP server utilizes OpenAI and a Vector store to interpret prompts and create Python code using PyAirbyte and context of the entire Airbyte Connector Registry for source and destination configuration parameters. If you want a complete write up of the architecture, check out our previous post .
The initial server was launched as a remote server with the goal to make installation as simple as possible, and allow us to ship quickly. The challenge is that the mcp spec is implemented differently in many tools. For remote servers, in particular, only Cursor supported passing in parameters as environment variables in the mcp.json. The TL;DR: was that if you wanted to use the server in any other IDE, you had to install it locally.
I’m excited to announce that, fast-pyairbyte has been updated to be an npx package which means it is now fully supported as a simple configuration install in your favorite AI tool. If it supports MCP, you can generate data pipelines with a single prompt.
Add fast-pyairbyte into your IDE or app. To add fast-pyairbyte, follow the instructions below for the most popular IDEs and tools. If you are using a different tool, all you will need is the same mcp.json server definition.
For all installations, you will need to add your own OpenAI API key into the env parameter. fast-pyairbyte uses the OpenAI Responses API to perform a natural language search.
Cursor Click the Install in Cursor button below for a single click setup.
Cline Click the manage MCP servers button , select the installed tab, then click Configure MCP Servers and paste in the mcp.json snippet with you OpenAI API Key:
Claude Desktop Go to Settings > Developer > Edit Config and paste in the mcp.json snippet with you OpenAI API Key:
Claude Code From the command line, execute:
You can verify the install via claude mcp list
Warp Open Settings > AI > Manage MCP servers . Select Add+ , and paste in the mcp.json snippet with you OpenAI API Key:
What’s next? The whole MCP and AI tooling space is rapidly changing. We are constantly looking at prompts, telemetry information, and community feedback to improve the service. One area we are actively looking at is improving performance as it takes approximately 60-90 seconds to generate the pipeline. The majority of this time is spent waiting for the vector search to complete, despite the amount of data being relatively Small. There is healthy discussion in the community whether Vector searches are even needed with MCP . Future releases of fast-pyairbyte may use a different approach as we optimize the service. For now, we will gather feedback and iterate. We’d love to hear your thoughts too , and don’t forget to join us for a YouTube Live event diving deeper into fast-PyAirbyte and other AI updates from Airbyte.
