AI agents use cursor_set_model to create or update resources in Vscode — usually the action step of a workflow, after the agent has gathered context. Every call changes real data in your Vscode environment.
This tool creates or modifies IDE configuration state (the active AI model selection) in a reversible manner. It does not execute arbitrary code, delete data, or perform destructive operations. While it affects which model processes subsequent requests, the action itself is a simple preference update with no side effects beyond configuration.
From the tool's definition Tool description states 'Change the active AI model in Cursor' — a reversible modification to IDE configuration state. The tool modifies a user preference setting without executing code, deleting data, or triggering external operations.
Attacks that exploit this kind of access
Change the active AI model in Cursor (e.g. claude-sonnet-4-5, gpt-4o, gemini-pro). It is categorised as a Write tool in the Vscode MCP Server, which means it can create or modify data. Consider rate limits to prevent runaway writes.
Register the Vscode MCP server in PolicyLayer and add a rule for cursor_set_model: 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 Vscode. Nothing to install.
cursor_set_model is a Write tool with medium risk. Write tools should be rate-limited to prevent accidental bulk modifications.
Yes. Add a rate_limit block to the cursor_set_model 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 cursor_set_model. 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.
cursor_set_model is provided by the Vscode MCP server (kloutdevs/vscode-mcp). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Every MCP server has a record like this.
Type a name, get the same breakdown: verified identity, auth posture, risk grade, capabilities, recommended policy.
Teams ship this data inside their own products. See what a licence covers →