AI Firewall MCP
This is a security layer for protecting LLM applications from adversarial inputs. It exposes five MCP tools over stdio: analyze_prompt runs your input through a three-agent pipeline (retrieval against FAISS vectors, guard signals, policy enforcement) to detect injection and jailbreak attempts. You also get get_threat_breakdown for per-signal scoring, sanitize_prompt to clean suspicious text, get_firewall_status for health checks, and benchmark_firewall to run the built-in adversarial test suite. Ships as a pip package or Docker container with configurable thresholds and operates in strict, moderate, or permissive modes. Reach for this when you're building multi-agent systems or user-facing LLM features and need a programmatic gate before prompts hit your model.
- tools
- 5
- transport
- STDIO
- last updated
- Jun 9, 2026
Tools
5 toolsPublic tool metadata for what this MCP can expose to an agent.
analyze_promptAnalyze a prompt for injection, jailbreaks, exfiltration, and leakage
Analyze a prompt for injection, jailbreaks, exfiltration, and leakage
No parameter schema in public metadata yet.
get_threat_breakdownDetailed per-signal scoring breakdown from the last analysis
Detailed per-signal scoring breakdown from the last analysis
No parameter schema in public metadata yet.
sanitize_promptClean a suspicious prompt while preserving legitimate content
Clean a suspicious prompt while preserving legitimate content
No parameter schema in public metadata yet.
get_firewall_statusHealth check: vector DB size, model status, uptime
Health check: vector DB size, model status, uptime
No parameter schema in public metadata yet.
benchmark_firewallRun the adversarial test suite and return detection statistics
Run the adversarial test suite and return detection statistics
No parameter schema in public metadata yet.
Sponsors
Agent CaffeineSponsoredClose your lid, come back to finished work
SnugSponsoredThe clipboard manager for the AI eraPromote your product hereReach 125k+ visitors monthly
WaxSponsoredBlock anything with supervision, without erasing your iPhonePromote your product hereReach 125k+ visitors monthly
knobsSponsoredSee your site on any phone, any day, in any languageReadme
<mcp-name: io.github.Akhilucky/ai-firewall-mcp>
AI Firewall — MCP Server
A multi-agent AI security layer that protects LLMs from prompt injection, jailbreaks, and policy violations. Available as an MCP server for any MCP-compatible client (Claude Desktop, Cursor, Windsurf, Cline, Roo Code, etc.).
Quick Start
Details
- Repository
- github.com/akhilucky/ai-firewall
- Website
- github.com/Akhilucky/AI-firewall
- Categories
- AI & LLM Tools, Security & Pentesting
- Package
ai-firewall-mcp- Transport
- STDIO
- Registry
- MCP registry, active
Related AI & LLM Tools MCP servers
View all
otgate — OT policy gatewayalikup-ai/otgateGates an LLM agent's writes to an OPC UA server on ranges, rate of change and live interlocks.0stars
Perspectivespolydev-ai/perspectives-mcpQuery GPT 5.2, Claude Opus 4.5, Gemini 3, Grok 4.1 simultaneously for AI perspectives0stars
ChiefLabbdentech/chiefmo-agentOpen execution contract for agents doing business work. 6 operators, one MIT-licensed spec.0stars
Hokmahdavidangularme/hokmah-mcp-serverAI Agent with Architectural Memory. Impact analysis (free), tests and code from the graph (pro).0stars
Mini Data Engine Runtime Copilotkroq86/data-engineering-runtime-labMCP runtime copilot for health checks, regressions, explainability, and operational memory.0stars
Fixflowmagneticdogson/fixflow-mcpCollective memory for AI agents. One agent solves a bug — every agent gets the fix instantly.0stars