An AI assistant that knows your pipelines

Most AI assistants are confidently generic. Ask one about your data platform and it will happily invent a table that does not exist. Ask BARRY is different for one structural reason: it is not guessing about your pipelines, it is connected to them.

The connection is an MCP server — an implementation of the Model Context Protocol that sits in front of the runtime and exposes it as a set of tools an assistant can actually call: list connections, browse tables, create a transport, get lineage, check client health, run a transport now, and more. When you ask which transports failed last night, the assistant does not summarise a hunch; it calls a tool, reads the real run history and answers from it. Ask it to move a table from SQL Server into Delta and it builds a transport through the same API a human would use.

That last point matters. The assistant acts through BARRY's own governed interfaces, not around them. Connections are capability-filtered, permissions still apply, and it can only do what the API allows — so a helpful assistant never becomes a backdoor. It reads and drives the platform under exactly the rules you already set.

There are two ways to meet it. Inside VS Code, the Ask BARRY chat panel answers questions in context and renders the results the way the product does — a transport as a transport, a chart as a chart, a lineage graph you can actually follow. And because the server speaks MCP, any MCP-capable client can connect to the same tools. The result is an assistant that has read your pipelines, knows which clients are online, and can trace a suspicious number back to the column it came from — because it is looking at the same system you are.

← All posts