Skip to main content
Back to AI Agents Hub
🛡️

Rebuff

by Protect AI

Self-hardening prompt injection detector

Rebuff is a specialized prompt injection detection framework that uses multi-layered defense including heuristics, LLM-based detection, vector database, and canary tokens.

Ease of Use
0/10
Community
0/10
Performance
0/10
Documentation
0/10

🎯 Key Features

Multi-layered detection

Heuristic analysis

LLM-based detection

Vector similarity search

Canary token injection

Self-hardening via feedback

API and SDK

Real-time detection

Confidence scoring

Attack pattern learning

Strengths

Specialized for prompt injection

Multi-layered approach

Self-improving system

Low latency

Easy integration

Good documentation

Active research backing

Limitations

Focused only on prompt injection

Requires API calls (hosted)

Free tier limitations

No output scanning

Limited language support

Best For

  • High-risk LLM applications
  • Customer-facing chatbots
  • Agent-based systems
  • RAG applications
  • Multi-tenant platforms

Not Recommended For