AI agents call graph_get to retrieve information from Meta without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
Although the tool can only read data (GET operation), the severity is elevated to 'high' rather than 'low' because: (1) it provides unrestricted access across the entire Meta Graph API surface, potentially exposing sensitive business intelligence, financial campaign data, and user information; (2) the 'escape-hatch' framing suggests it bypasses normal safety constraints; (3) misuse could leak confidential marketing…
From the tool's definition The tool is described as a 'raw GET to any Graph API endpoint' with no modification capability. GET requests are read-only operations. The server is explicitly marked as 'Read-only MCP server'.
Attacks that exploit this kind of access
Escape-hatch: raw GET to any Graph API endpoint. Path must start with. It is categorised as a Read tool in the Meta MCP Server, which means it retrieves data without modifying state.
Register the Meta MCP server in PolicyLayer and add a rule for graph_get: 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 Meta. Nothing to install.
graph_get 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 graph_get 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 graph_get. 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.
graph_get is provided by the Meta MCP server (nourpups/meta-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 →