The Industry-Standard Deterministic Security Layer for AI Agents.
SupraWall solves the "trust" problem in generative AI by providing a zero-knowledge, deterministic security layer between LLMs and your production systems. Unlike probabilistic security models, SupraWall enforces strict business policies and forensic-grade threat detection before an agentic action even reaches your API.
Official Website | Main Repository | Documentation
SupraWall (formerly AgentGate) was built to bridge the gap between high-velocity AI experimentation and enterprise security compliance. Our goal is to empower developers to ship autonomous agentic fleets without the risk of Prompt Injection, PII data leakage, or uncontrolled tool execution.
This monorepo serves as the official distribution hub for all SupraWall plugins and framework integrations.
- Dify Plugin: Explicit security guardrails for Dify Workflows.
- MCP Claude Server: Official implementation of the Model Context Protocol (MCP).
- AWS Marketplace Integration: Pre-certified Guardrail tool for AWS Cloud (SaaS & Container).
- OpenClaw Extension: Native security middleware for the OpenClaw agent runtime.
Seamless "Zero-Trust" wrappers for the industry's leading AI agent frameworks:
- AutoGen Integration: Secure protection for Microsoft AutoGen multi-agent topologies.
- CrewAI Integration: Deterministic policy enforcement for CrewAI processes and agents.
- LangChain (Python & TS): Security interceptors for LangChain chains, tools, and executors.
- LlamaIndex Integration: Secure guardrails for RAG pipelines and tool-based retrieval.
- Vercel AI SDK Integration: One-line security additions for projects built on the Vercel AI SDK.
We represent the most robust security option on the Dify Marketplace. To install manually during the initial review phase, use the pre-packaged .difypkg found in the suprawall-security directory.
Add SupraWall to your claude_desktop_config.json:
{
"mcpServers": {
"suprawall": {
"command": "npx",
"args": ["-y", "@suprawall/mcp-plugin"]
}
}
}SupraWall is built on three core pillars:
- Detection: Real-time identification of adversarial intent (Prompt Injection, SQLi, Jailbreaks).
- Scrubbing: Automatic removal of PII (Names, Emails, Keys) before data leaves your environment.
- Enforcement: Strict policy-based execution—if a tool call doesn't match your rule, it's blocked.
Licensed under the Apache License 2.0.
© 2026 WiserAutomation. Agency. All rights reserved. Registered in the Dify and Anthropic Marketplaces.