Check the content around an agent before it becomes an action.
Find risky structure and likely secrets, redact sensitive patterns, validate model output, or run the full preparation pipeline. These tools return review signals and deterministic results—not false guarantees.
Scan untrusted text
Flag structural obfuscation, suspicious instruction patterns, mixed scripts, and encoded payload signals.
Scan text →Detect secrets
Find likely API keys, private keys, JWTs, cloud credentials, and related patterns without returning their values.
Detect secrets →Redact sensitive text
Replace common credential and personal-identifier patterns while preserving the rest of the text.
Redact text →Validate AI output
Parse JSON and enforce common JSON Schema rules before an application acts on model output.
Validate output →Sanitize agent context
Normalize untrusted context, optionally redact sensitive patterns, and return deterministic risk indicators.
Sanitize context →Prepare for agent
Run normalization, inspection, risk scanning, secret detection, and optional redaction in one call.
Prepare text →Choose the narrowest check
Use one focused tool when you know the risk you are testing. Use Prepare for agent when a workflow needs a consistent intake pipeline covering several common checks in one request.
Keep decisions outside the model
The API and MCP tools return structured evidence your application can evaluate. They do not decide whether content is safe, authorize an action, or replace application-specific controls.