browser_eval · Claude Flow MCP: Risk & Policy

WHAT IT DOES

What browser_eval does on Claude Flow

AI agents invoke browser_eval to trigger actions in Claude Flow. 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.

WHY IT NEEDS A POLICY

Why browser_eval needs a policy

browser_eval executes arbitrary JavaScript in a live browser context, which can trigger external operations, modify page state, steal cookies/credentials during login flows, and interact with third-party services. While not destructive or financial by itself, the capability to automate browser actions and manipulate DOM elements represents a high-severity Execute risk.

From the tool's definition Tool description explicitly states 'Execute JavaScript in page context' and mentions 'browser automation', 'login flows', and 'replay against DOM'. These are hallmarks of code execution with external side effects.

Attacks that exploit this kind of access

Recommended rule Ready

Verdict

browser_eval Rate-limited

browser_eval 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.

View as policy code

policy.json

{
  "version": "1",
  "default": "deny",
  "tools": {
    "browser_eval": {
      "limits": [
        {
          "counter": "browser_eval_rate",
          "window": "minute",
          "max": 10,
          "scope": "grant"
        }
      ]
    }
  }
}

RATE-LIMIT THIS TOOL →

Instant setup, no code required.

PolicyLayer is an MCP gateway: every tool call is checked against your rule before it reaches Claude Flow, so you can run agents against it safely.

FAQ

Questions about browser_eval

What does the browser_eval tool do? +

Execute JavaScript in page context Use when native WebFetch is wrong because you need real browser automation — JS-heavy SPA scraping, login flows with cookie reuse, replay against DOM-drifted versions, AIDefence PII gating before content reaches Claude. For static HTML pages, native WebFetch is faster and free. It is categorised as a Execute tool in the Claude Flow MCP Server, which means it can trigger actions or run processes. Use rate limits and argument validation.

How do I enforce a policy on browser_eval? +

Register the Claude Flow MCP server in PolicyLayer and add a rule for browser_eval: 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 Claude Flow. Nothing to install.

What risk level is browser_eval? +

browser_eval is a Execute tool with high risk. Execute tools should be rate-limited and have argument validation enabled.

Can I rate-limit browser_eval? +

Yes. Add a rate_limit block to the browser_eval 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.

How do I block browser_eval completely? +

Set action: deny in the PolicyLayer policy for browser_eval. 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.

What MCP server provides browser_eval? +

browser_eval is provided by the Claude Flow MCP server (claude-flow). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.