AI-Powered Tool Generation
Autonomous, self-extending MCP server that generates Python tools on-demand with enterprise-grade security auditing. Request tools in natural language, get production-ready, sandboxed code.
From natural language request to production-ready tool in four automated steps.
Describe the tool you need in natural language via MCP protocol from Claude, GPT, or any AI assistant.
Engineer Agent creates Python code using your configured LLM with sandbox-safe patterns.
Guardian Agent performs Bandit SAST static analysis + LLM contextual review for vulnerabilities.
Human approval via dashboard, then tool is live and available via MCP.
Four specialized agents work together to generate secure, tested tools.
Orchestrator
Coordinates workflow, handles retries (max 3 attempts), and queues approved tools for deployment.
Code Generator
Generates Python code via LLM using optimized prompts for sandbox-safe, production-ready output.
Security Auditor
Two-phase audit: Bandit SAST static analysis + LLM contextual review catches vulnerabilities others miss.
Validator
Validates code executes correctly in Docker sandbox with strict resource limits.
Clean, intuitive interface for tool generation and management.
Request tools in natural language. MCP-connected for seamless AI integration.
Real-time agent activity. Watch Engineer, Guardian, and Tester agents work.
Enterprise-grade features for secure AI tool generation.
Combines Bandit SAST static analysis with LLM contextual review. Catches both pattern-based vulnerabilities and semantic security issues that traditional scanners miss.
All generated tools run in isolated Docker containers with strict resource limits, no network access, and 5-second timeout. Complete isolation from your infrastructure.
Native Model Context Protocol support (2024-11-05+ Streamable HTTP spec) for seamless integration with Claude, GPT, and other AI assistants.
Each authenticated user has their own isolated tool namespace and API keys. Per-user Redis namespacing ensures complete data separation.
Run Guardian Forge entirely within your infrastructure. Docker Compose deployment, your LLM keys, your data stays yours. Air-gapped environment support.
Works with OpenAI, Anthropic, Google Gemini, Azure OpenAI, Ollama (local), and OpenRouter. Hot-swap providers without restart.
Simple natural language requests create production-ready tools.
Guardian Forge generates, audits, tests, and deploys a secure weather tool automatically.
Self-hosted deployment with your own LLM keys. Complete control over your AI tool generation infrastructure.
github.com/virtuallifehub-del/guardian-forge-dist
