AI agents call web_fetch to retrieve information from Mcp Nexus without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
This is fundamentally a Read operation: it retrieves content from a URL and converts it to markdown format. No data is created, modified, or deleted. Severity is medium rather than low because fetching arbitrary URLs could expose sensitive content (credentials in logs, private documents, SSRFs if the URL is attacker-controlled), and a malicious agent might use URL fetching to reconnaissance or exfiltrate information…
From the tool's definition Tool fetches and extracts markdown content from a URL—a retrieval operation with no modification or destruction. Description explicitly states 'Fetch and extract' indicating data retrieval.
Attacks that exploit this kind of access
Fetch and extract markdown content from a URL with provider fallback. It is categorised as a Read tool in the Mcp Nexus MCP Server, which means it retrieves data without modifying state.
Register the Mcp Nexus MCP server in PolicyLayer and add a rule for web_fetch: 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 Mcp Nexus. Nothing to install.
web_fetch 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 web_fetch 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 web_fetch. 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.
web_fetch is provided by the Mcp Nexus MCP server (xydong-web/mcp-nexus). 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 →