Use a mouse and keyboard to interact with a web browser, and take screenshots.
AI agents invoke computer to trigger actions in OpenChrome. 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 controls a real Chrome browser via mouse and keyboard, meaning an AI agent can navigate to any URL, click buttons, submit forms, and perform authenticated actions. The effects are entirely dependent on what the agent instructs it to do, making it an Execute-category tool.
From the tool's definition "Use a mouse and keyboard to interact with a web browser" — triggers real browser actions (clicks, keystrokes, navigation) and "take screenshots" implies live browser control with arbitrary side effects depending on agent inputs.
Documented attack patterns abuse exactly the kind of access computer gives an agent:
PolicyLayer is an MCP gateway — it sits between your AI agents and OpenChrome, and nothing reaches the server without passing your rules. This is the rule we recommend for computer:
{
"version": "1",
"default": "deny",
"tools": {
"computer": {
"limits": [
{
"counter": "computer_rate",
"window": "minute",
"max": 10,
"scope": "grant"
}
]
}
}
}computer 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.
Use a mouse and keyboard to interact with a web browser, and take screenshots. It is categorised as a Execute tool in the OpenChrome MCP Server, which means it can trigger actions or run processes. Use rate limits and argument validation.
Register the OpenChrome MCP server in PolicyLayer and add a rule for computer: 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 OpenChrome. Nothing to install.
computer 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 computer 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 computer. 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.
computer is provided by the OpenChrome MCP server (shaun0927/openchrome). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Deterministic rules across all 106 OpenChrome tools. Per-identity grants. Full audit log. Live in minutes. Nothing to install.
Free to start. No card required.
106 OpenChrome tools catalogued and risk-classified — across an index of 42,500+ MCP servers.