AI agents call token_savings_report to retrieve information from Paparats without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
This tool performs read-only analysis of token consumption metrics by comparing different approaches (naive baseline, search-only, actual consumption). It correlates chunk_fetches data to generate a report. No side effects, no code execution, no data modification or deletion occurs. It is a reporting/analytics tool with no blast radius if misused by an AI agent—worst case being that it returns incorrect statistics.
From the tool's definition Tool name 'token_savings_report' and description 'Token-savings estimates: naive baseline vs. search-only vs. actually-consumed' indicate the tool retrieves and reports statistics/metrics.
Attacks that exploit this kind of access
Token-savings estimates: naive baseline vs. search-only vs. actually-consumed (uses chunk_fetches correlation). It is categorised as a Read tool in the Paparats MCP Server, which means it retrieves data without modifying state.
Register the Paparats MCP server in PolicyLayer and add a rule for token_savings_report: 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 Paparats. Nothing to install.
token_savings_report 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 token_savings_report 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 token_savings_report. 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.
token_savings_report is provided by the Paparats MCP server (@paparats/cli). 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 →