Think inside your AI world.
How do I connect Databricks to Claude?
Databricks ships managed MCP servers over Unity Catalog, AI Search, Genie spaces, custom functions and AI Gateway SaaS tools — the lakehouse, conversationally addressable.
Direct answer: connect Databricks' managed MCP servers in Claude — Settings → Connectors, authorise the workspace ('claude mcp add' in Claude Code). The surface spans Unity Catalog (governed data assets), AI Search, Genie spaces for natural-language analytics, custom functions and AI Gateway tools. Claude asks; the lakehouse answers within the governance Unity Catalog already enforces.
The lakehouse answers in place
What assets exist for the customer domain, what the Genie space returns for a revenue question, what the custom function computes — resolved through managed servers that inherit workspace governance rather than bypass it.
The bar a pipeline never sets for itself
Worked case: your team ratified a retraining bar — the churn model retrains only when the upstream tables pass freshness and completeness checks, set after a model once retrained on a half-loaded week and spent a month being confidently wrong. Databricks runs the checks and the retrain; that the second waits on the first is your operating law.
Pipeline reads with the law standing
Through Unl the retrain conversation arrives framed: completeness check failing on two tables — the bar says wait, and the confidently-wrong month is why the bar exists. The lakehouse executes; Unl holds the law the execution answers to; the read stays one-way throughout.
Databricks to Claude reads the governed lakehouse; Databricks through Unl reads it under the operating laws you ratified — the retraining bar arrives before the retrain button does.
Reads through Unl arrive with measured context — in the presence of the decisions you’ve already settled. The reach lane is live: one box, paste anything. If it speaks MCP, Unl can reach it. Readings arrive unprompted, the data beside the criterion; Unl is a courier, not a warehouse, and keeps only your keys and the frame.
Read further
Questions people ask
What do the managed Databricks MCP servers expose?
Unity Catalog assets, AI Search, Genie spaces, custom functions and AI Gateway SaaS tools — managed by the platform, inheriting the workspace's existing governance.
Is the connection portable to other MCP clients?
Yes; managed servers speak the standard protocol. Other clients — ChatGPT included — connect via their own connector menus with the same authorisation.
How is Unl different from Unity Catalog governance?
Unity Catalog governs the data: who reads what. Unl holds your judgement about operations over the data: when a model may retrain, which checks gate which actions, and why. Different layers; they compose.
What if the model degrades faster than the retraining bar allows?
Then the bar is wrong and the readings that forced the question are the evidence for changing it. A threshold set before you knew the decay curve is a first guess; what matters is that replacing it is a deliberate act with the old reasoning still visible, not a quiet edit to a number.
What this is
Think inside your AI world — you stay in command
Save the thoughts, decisions and targets worth keeping, each with its reasoning, carried into every AI session the moment they matter. A new unit of exchange between you and your AI: the Settled Why with standing that travels. Unprompted.
Your whole AI world. What you decided at the epicentre. It plugs into Claude, Claude Code, ChatGPT and Cursor as an MCP connector — quick to connect, in a couple of steps.
MCP native·Human settled·Model agnostic·Your data
Measured Context
Connect Unl to bring the right information into the moment.
Your sources, read against the criteria you set.
The full product, open. Free at launch.