Sanitize Untrusted Context Before an AI Agent Reads It
Normalize hidden/control characters, return structural risk signals, detect secrets, and optionally redact sensitive patterns before untrusted context reaches a model.
When to use it
Use this as a preprocessing step for email, webpages, document extraction, tickets, retrieval results, or other external text. It preserves semantic text while removing mechanical artifacts and can redact sensitive patterns.
Important limitation
It does not remove or neutralize every semantic prompt-injection instruction and does not certify that content is safe.
Use it in your browser
The browser version runs locally on your device.
Open the tool →Use it from an agent or application
POST https://api.scrubmytext.com/v1/sanitize-context
MCP tool name: sanitize_agent_context. Paid API tools use a bearer API key; inspect_text is also available on the free API tier.