AI agents call browser_screenshot to retrieve information from MCProxy without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
Taking a screenshot retrieves visual information from the current browser state with no side effects, data modification, or external operations triggered. It is purely observational. While the broader MCProxy server enables browser automation with potential for misuse, this specific tool is constrained to data retrieval.
From the tool's definition Tool name is 'browser_screenshot' and description states it 'Take[s] a screenshot of the current page. Returns base64-encoded PNG image.' This is a read-only operation that captures visual state without modifying data or triggering side effects.
Attacks that exploit this kind of access
Take a screenshot of the current page. Returns base64-encoded PNG image. Optionally saves to a local file. It is categorised as a Read tool in the MCProxy MCP Server, which means it retrieves data without modifying state.
Register the MCProxy MCP server in PolicyLayer and add a rule for browser_screenshot: 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 MCProxy. Nothing to install.
browser_screenshot is a Read tool with low risk. Read-only tools are generally safe to allow by default.
Yes. Add a rate_limit block to the browser_screenshot 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 browser_screenshot. 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.
browser_screenshot is provided by the MCProxy MCP server (saladtechnologies/mcproxy). 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 →