AI agents call search_clinical_trials to retrieve information from Pgx without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
Clinical trial searches are typically read-only queries against medical databases. There is no indication this tool modifies data, executes code, deletes records, or commits financial transactions. The worst-case misuse would be retrieving inappropriate trial information, which has low blast radius compared to other risk categories.
From the tool's definition Tool name 'search_clinical_trials' indicates a query/search operation. The server context describes 'analysis' and 'connecting to' databases (ClinVar, PharmGKB, gnomAD), which are read-only lookups.
Attacks that exploit this kind of access
search_clinical_trials. It is categorised as a Read tool in the Pgx MCP Server, which means it retrieves data without modifying state.
Register the Pgx MCP server in PolicyLayer and add a rule for search_clinical_trials: 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 Pgx. Nothing to install.
search_clinical_trials 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_clinical_trials 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_clinical_trials. 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_clinical_trials is provided by the Pgx MCP server (julius-schmidt/mcp-pharmacogenomics). 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 →