LLM security guard API

Rakshak sits between your application and your LLM. Use the input guard before forwarding user text to the model, the output guard before showing model text to users, and the conversation guard when risk builds across multiple turns.

• Prompt injection, jailbreaks, obfuscation, and data extraction attempts.
• PII, system prompt leaks, internal URLs, and unsafe model responses.
• Multi-turn escalation, payload splitting, and repeated probing.

All production guard endpoints use an API key in the X-API-Keyheader. Customer API keys use the rsk_ prefix.

curl -X POST "/v1/scan" \
  -H "Content-Type: application/json" \
  -H "X-API-Key: rsk_your_key" \
  -d '{"prompt": "Ignore previous instructions"}'

import rakshak

client = rakshak.Client(
    api_key="rsk_your_key",
)

result = client.guard("user message here")
if result.blocked:
    return "I can't help with that."

# safe: forward the original user message to your LLM