Databricks

Connect AI agents to the Databricks REST API. Agents run SQL on a warehouse, orchestrate jobs, manage clusters, browse Unity Catalog, and inspect DBFS and workspace paths.

24 actions available

A scheduled workload needs compute and a SQL result

Agent runs Start SQL Warehouse and polls Get SQL Warehouse until it is running

Agent runs Execute SQL Statement to compute the metric on the warehouse

If the statement is still running at the wait timeout, agent polls Get Statement Result

Agent posts the result to the team via Slack

Agent runs Stop SQL Warehouse to release compute when the workload finishes

Any Run Job Now or Terminate Cluster against a shared production resource pauses for approval

What This Integration Enables

Agents run parameterized SQL on a warehouse and poll long statements to completion, trigger existing jobs with parameter overrides and track each run to a terminal state, start and stop warehouses and clusters to control cost, and browse Unity Catalog, DBFS, and workspace paths for data-asset discovery. Execution can be synchronous within a wait timeout, or asynchronous, in which case Execute SQL Statement returns a `statement_id` and the agent polls Get Statement Result until the state is terminal. Warehouse and cluster start and stop calls return immediately, so the agent polls Get SQL Warehouse or Get Cluster to confirm the transition.

Without FlowRunner

Compute always on A warehouse left running around a scheduled job burns cost

Jobs triggered by hand Someone opens the workspace to run a job and watch it

SQL results stay in notebooks Query output never reaches the systems that could act on it

With FlowRunner

Compute on demand The flow starts a warehouse for the workload and stops it after

Jobs triggered in the flow Run Job Now kicks off a job and tracks it to a terminal state

Results move onward Execute SQL Statement feeds the flow and posts to channels

Use Case Scenarios

Cost-aware scheduled query

Before a scheduled query, the agent calls Start SQL Warehouse and polls Get SQL Warehouse until it is running. It runs Execute SQL Statement, posts the result to the team channel with Slack, then calls Stop SQL Warehouse to release compute. The warehouse is only awake for the work that needs it.

Job orchestration with tracking

The agent calls Run Job Now with parameter overrides, then polls Get Run until the run reaches a terminal state, and records the outcome downstream. A job that used to require someone watching the workspace now runs and reports on its own.

Shared-resource action with a human gate

An agent is about to trigger a heavy production job or terminate a shared cluster. Before it runs Run Job Now or Terminate Cluster on a resource flagged as shared production, it does not act on its own. It routes the action for approval and proceeds only after a person confirms.

Human-in-Loop Highlight

On Databricks the expensive mistakes are not reads; they are actions on shared compute, like triggering a heavyweight production job or terminating a cluster other workloads depend on. FlowRunner's answer is human-in-the-loop, an execution pattern where the agent pauses on its own, assembles the context and the choices, routes to a human on their preferred channel, and resumes the moment they respond. When a flow reaches Run Job Now for a job flagged as production, or Terminate Cluster on a shared cluster, the agent pauses and asks through Slack: "Ready to Run Job Now for `nightly-etl` with these parameter overrides. Approve or cancel?" The action runs only after a person confirms, with the approver and timestamp captured in the run log. A connector can trigger any job; an orchestration layer knows which triggers should stop and ask.

Agent processes routinely

Detects exception requiring judgment

Clear match Continues automatically

Ambiguous Routes to human via Slack

Human decides

Agent resumes with decision

Agent Capabilities

24 actions

SQL

Execute SQL Statement Runs a parameterized SQL statement on a SQL warehouse, synchronously or asynchronously depending on the wait timeout, and returns the result manifest and inline data or a `statement_id` to poll. Bind values with named `:name` parameters.
Get Statement Result Polls a running statement by id until it reaches a terminal state and returns the result.
Cancel Statement Aborts a running SQL statement by id.
List SQL Warehouses Lists the SQL warehouses in the workspace with their state.
Get SQL Warehouse Returns the current state of one SQL warehouse, for confirming it is running.
Start SQL Warehouse Starts a stopped SQL warehouse. The transition is asynchronous; poll Get SQL Warehouse to confirm.
Stop SQL Warehouse Stops a running SQL warehouse to control compute cost.

Jobs

List Jobs Lists the jobs defined in the workspace.
Get Job Returns the definition and settings of one job.
Run Job Now Triggers an existing job on demand with optional parameter overrides and returns the run id to track.
List Job Runs Lists recent runs of a job with their states.
Get Run Returns the status of one job run; poll it until the run reaches a terminal state.
Get Run Output Returns the output of a completed job run.
Cancel Run Cancels a running job run.

Clusters

List Clusters Lists the clusters in the workspace with their state.
Get Cluster Returns the current state of one cluster.
Start Cluster Starts a terminated cluster. The transition is asynchronous; poll Get Cluster to confirm.
Terminate Cluster Terminates a running cluster to control compute cost.

Unity Catalog

List Catalogs Lists the Unity Catalog catalogs available to the workspace.
List Schemas Lists the schemas in a catalog.
List Tables Lists the tables in a schema, so an agent can find data before querying.

Files

List DBFS Path Lists the contents of a DBFS path.
List Workspace Path Lists the contents of a workspace path, for browsing notebooks and folders.

Account

Get Current User Returns the identity the API token belongs to, useful as a quick connection check.

Start building with Databricks

$100 in credits. No card required. Connect in minutes.

Start Building Free Talk to us