AI agents invoke start_cluster to trigger actions in Databricks MCP Server. What it does depends on the arguments the agent supplies, and its effects often reach beyond the immediate call — builds kicked off, notifications sent, workflows started.
This tool executes an action that launches compute infrastructure with side effects determined by the tool's arguments (cluster ID). While it does not permanently delete or move money directly, it triggers external systems and can incur costs. It is not merely Read (no data retrieval), nor Write (not creating/modifying reversible data in a database sense), nor Destructive (cluster can be re-terminated).
From the tool's definition Tool description states 'Start a terminated cluster' — this triggers an external operation (cluster startup) whose effects depend on which cluster is targeted.
Documented attack patterns abuse exactly the kind of access start_cluster gives an agent:
PolicyLayer is an MCP gateway — it sits between your AI agents and Databricks MCP Server, and nothing reaches the server without passing your rules. This is the rule we recommend for start_cluster:
{
"version": "1",
"default": "deny",
"tools": {
"start_cluster": {
"limits": [
{
"counter": "start_cluster_rate",
"window": "minute",
"max": 10,
"scope": "grant"
}
]
}
}
}start_cluster stays usable, but rate-capped — a runaway agent can't fire it dozens of times a minute. Everything else on the server is denied unless you say otherwise.
Free to start. No card required.
Start a terminated cluster. It is categorised as a Execute tool in the Databricks MCP Server MCP Server, which means it can trigger actions or run processes. Use rate limits and argument validation.
Register the Databricks MCP Server MCP server in PolicyLayer and add a rule for start_cluster: allow, deny, rate-limit, or require approval. Point your MCP client at the PolicyLayer proxy URL and the rule is enforced on every call, before it reaches Databricks MCP Server. Nothing to install.
start_cluster is a Execute tool with high risk. Execute tools should be rate-limited and have argument validation enabled.
Yes. Add a rate_limit block to the start_cluster rule in your PolicyLayer policy. For example, setting max: 10 and window: 60 limits the tool to 10 calls per minute. Rate limits are tracked per agent session and reset automatically.
Set action: deny in the PolicyLayer policy for start_cluster. The AI agent will receive a policy violation error and cannot call the tool. You can also include a reason field to explain why the tool is blocked.
start_cluster is provided by the Databricks MCP Server MCP server (markov-kernel/databricks-mcp). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Start from Databricks MCP Server, add the rest of your stack, and see everything your agents can call. Then put policy on all of it.
Free to start. No card required.
38 Databricks MCP Server tools catalogued and risk-classified — across an index of 43,000+ MCP servers.