AI agents call search_web to retrieve information from Web Fetch without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
This tool retrieves and queries data (search results) with no side effects. It does not create, modify, delete, or execute operations. The action is purely informational retrieval, fitting the Read category. Severity is low as misuse would only expose publicly available search results.
From the tool's definition Tool performs web search via Bing search engine and returns search results list. Description translates to 'Search keywords using Bing search engine, return search results list.' No modification, deletion, or execution of code/commands occurs.
Attacks that exploit this kind of access
使用 Bing 搜索引擎搜索关键词，返回搜索结果列表. It is categorised as a Read tool in the Web Fetch MCP Server, which means it retrieves data without modifying state.
Register the Web Fetch MCP server in PolicyLayer and add a rule for search_web: 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 Web Fetch. Nothing to install.
search_web 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 search_web 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 search_web. 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.
search_web is provided by the Web Fetch MCP server (xiaozhuabcd1234/web-fetch-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 →