How Can AI Agents Screen Text for Prompt Injection and PII?

A developer can use Exepal to call tools that scan text for prompt injection, secrets, and personally identifiable information (PII). An agent can discover available tools through MCP or Exepal’s tools endpoint, review each tool’s description and price, then call the relevant tool and receive a JSON response.

Find and call a screening tool

Exepal lists text-scanning tools for prompt injection, secrets, and PII. Its documented workflow is for the agent to list the available tools, inspect what each one does and costs, and call a suitable tool. The product information says each call returns JSON, but it does not provide the screening tools’ specific schemas or output fields.

For example, an agent workflow could send text it has retrieved from an external source to a screening tool before passing that text into a later step. This describes a possible workflow, not a guarantee that the scan will identify every issue.

Understand how calls are paid for

Exepal says calls can use prepaid credit, x402 with USDC, or free daily calls. The service states that it offers 20 free calls per day without requiring a credit card or wallet. Its product information does not give the individual prices for the screening tools, so developers should check the tool listing for current details.

Treat screening as a check, not proof

The available product information does not describe detection methods, accuracy, input limits, or guarantees. A successful response therefore should not be treated as proof that text is free of prompt injection, secrets, or PII. Developers should account for those unknowns when deciding how to handle text that may be sensitive or untrusted.

Exepal is aimed at developers building agent workflows who want callable text-screening tools through MCP or an API. To assess whether a specific tool fits your workflow, review its description and price in Exepal’s tool listing before making a call.