Content safety
One place for the tools that clean, inspect, protect, validate, and prepare content. Humans get focused browser workflows; agents keep precise, composable API contracts.
Clean and understand content
Normalize representation, reveal hidden structure, compare versions, or rework wording.
ScrubMyText
Remove hidden controls and normalize copied text without rewriting meaning.
InspectMyText
See exact Unicode, zero-width, bidirectional, and mixed-script anomalies.
CompareMyText
Verify exact changes and normalization-only differences.
ReworkMyText
Deterministically revise wording without a language model.
ScanMyDocument
Inspect documents locally for hidden content, metadata, macros, and suspicious structure.
Protect agent inputs and outputs
Use focused tools when one safeguard needs to be audited, or the combined preparation pipeline for a default boundary.
ScanMyText
Flag structural obfuscation and suspicious instruction patterns.
DetectMySecrets
Find likely credentials without returning secret values.
RedactMyText
Replace common credential and personal-identifier patterns.
ValidateMyOutput
Validate JSON and common schema rules before software acts.
SanitizeMyContext
Normalize and optionally redact untrusted agent context.
PrepareMyContext
Combine normalization, inspection, risk checks, secret detection, and optional redaction.
Choose the check by failure mode
“Clean this text” can mean several different things. Start with the risk you need to observe or control.
Representation changed
Use ScrubMyText when copied text contains zero-width characters, unusual spacing, bidirectional controls, or inconsistent Unicode forms. Use InspectMyText when you need a report before changing anything, and CompareMyText when you must prove exactly what differed.
Content crossed a trust boundary
Use ScanMyText for structural obfuscation and instruction-like patterns, DetectMySecrets for likely credentials, and RedactMyText for supported sensitive-value patterns. A finding is a signal for your policy; absence of a finding is not proof that content is benign.
Software will act on the result
Use ValidateMyOutput to enforce JSON shape and supported schema rules. Use PrepareMyContext when an agent needs one ordered pipeline that normalizes, inspects, scans, detects, and optionally redacts before downstream processing.
A reproducible safety boundary
A practical sequence for emails, retrieved documents, pasted text, webhook payloads, and other content your application did not author.
Keep the original outside the prompt
Retain a hash or protected copy before transformation when auditability matters. Do not put a live credential into a diagnostic request merely to test whether the detector can find it.
Run the narrowest checks
Normalize representation, review reported code points and risk flags, detect secret categories, then apply explicit redaction or rejection rules. REST responses and MCP tool results use the same focused contracts.
Let your application enforce policy
Route high-risk findings to review, reject invalid structured output, or send the prepared value downstream. Store the tool version and result digest if the decision must later be explained.
Local and server processing
The browser document scanner performs supported file inspection locally so the document does not need to be uploaded for that workflow. API and MCP calls are processed by the ScrubMyText service. Check each tool's page and the privacy policy before sending sensitive material.
Important limits
These tools are not antivirus software, a complete data-loss-prevention system, a legal compliance certification, or a semantic guarantee against prompt injection. File formats, encodings, schemas, and detection patterns have documented boundaries. Layer them with least-privilege access and application-specific authorization.