A Protocol and Service for Building Great and Secure Software with AI Agents
Guided.dev provides the missing link between AI coding assistants and best-practice software development: a curated knowledge graph of secure coding patterns, vulnerabilities, and architectural guidance, accessible via a standardized protocol.
Guided.dev consists of two key innovations:
-
A Curated Knowledge Graph: A security-first knowledge base built on PostgreSQL with Apache AGE, containing:
- Technologies and frameworks
- Security vulnerabilities (OWASP Top 10)
- Best practices and secure coding patterns
- Deployment strategies
- Architectural guidance
-
The AGENTS.md Protocol: A machine-readable specification that tells AI agents how to discover and interact with our guidance service, enabling structured, secure, and expert-informed development.
Phase 0 (Foundation & Setup) - Complete ✓ Phase 1 (Admin CRUD Interface) - Complete ✓ Phase 2 (MCP Server) - Complete ✓
- ✓ PostgreSQL with Apache AGE graph database
- ✓ Graph query interface (
Guided.Graph) - ✓ Initial knowledge base (24 nodes, 24 relationships)
- ✓ Custom mix tasks for setup and seeding
- ✓ Admin CRUD interface with Phoenix LiveView
- ✓ Knowledge graph management UI
- ✓ MCP Server with three core tools:
tech_stack_recommendation- Get tech stack advice for use casessecure_coding_pattern- Retrieve secure code patternsdeployment_guidance- Get deployment recommendations
- ✓ AGENTS.md protocol specification
Prerequisites: Erlang/OTP 28+, Elixir 1.19.1+, Docker
# Clone the repository
git clone https://github.com/carverauto/guided.git
cd guided
# Run the automated setup script
./scripts/dev_setup.sh
# Start the Phoenix server
mix phx.serverVisit localhost:4000 from your browser.
For detailed setup instructions and manual installation, see docs/DEVELOPMENT.md.
Connect Claude Desktop to guided.dev to get AI-powered secure coding guidance in your conversations!
Quick Setup:
- Start the server:
mix phx.server - Add to your
claude_desktop_config.json:{ "mcpServers": { "guided-dev": { "command": "npx", "args": ["mcp-remote", "http://localhost:4000/mcp"] } } } - Restart Claude Desktop
- Ask Claude: "Do you have access to guided.dev?"
What you get:
- 🔒 Security-first recommendations with OWASP Top 10 advisories
- 💻 Secure code patterns with real examples
- 🚀 Deployment guidance for your tech stack
📖 Full guide: CLAUDE_DESKTOP_SETUP.md
LLMs are powerful code generators but they operate in a vacuum. They often produce code that is functional but naive in its architecture, security, and scalability. There's no standardized way for AI agents to discover and consume curated best-practice knowledge.
Guided.dev provides:
- A trusted source of security-first, opinionated guidance
- A clear protocol (AGENTS.md) for AI agents to consume this guidance
- A graph-based knowledge model for complex relationships between technologies, vulnerabilities, and mitigations
An AI agent building a Python web app:
- Discovers the project's
AGENTS.mdfile - Queries guided.dev's MCP server for tech stack recommendations
- Receives guidance on Streamlit + SQLite with security advisories
- Queries for secure coding patterns specific to those technologies
- Generates code using parameterized queries and proper input validation
- Backend: Elixir / Phoenix Framework
- Database: PostgreSQL with Apache AGE extension (graph database)
- Frontend: Phoenix LiveView
- Query Language: openCypher
- API: Model Context Protocol (MCP)
Phase 0 (Foundation):
- ✓ PostgreSQL with Apache AGE graph database
- ✓ openCypher query interface
- ✓ Initial knowledge base covering:
- Python, Streamlit, SQLite, FastAPI
- OWASP Top 10 vulnerabilities
- Security controls and mitigations
- Best practices and deployment patterns
Phase 1 (Admin Interface):
- ✓ LiveView-based admin CRUD interface
- ✓ Node and relationship management
- ✓ Knowledge graph dashboard
- ✓ Graph visualization
Phase 2 (MCP Server):
- ✓ Public-facing MCP API at
/mcpendpoint - ✓ Three core tools (capabilities):
tech_stack_recommendation- Opinionated tech stack advice with security advisoriessecure_coding_pattern- Secure code snippets and best practicesdeployment_guidance- Deployment pattern recommendations
- ✓ AGENTS.md protocol specification for AI agent discovery
- ✓ Integration with Hermes MCP library
- ✓ Streamable HTTP transport
- Expanded knowledge domains (JavaScript, Go, Rust, etc.)
- Community contribution system
- IDE integrations
- Advanced graph analytics and insights
For detailed setup instructions, see docs/DEVELOPMENT.md.
# Run tests
mix test
# Format code
mix format
# Full pre-commit check (compile, format, test)
mix precommit
# Reset database and graph
mix ecto.reset
mix graph.setup
mix graph.seed
# Start Phoenix server
mix phx.server
# Start with IEx (interactive Elixir)
iex -S mix phx.server# In IEx
iex> Guided.Graph.query("MATCH (t:Technology) RETURN t.name")
{:ok, [["Python"], ["Streamlit"], ["SQLite"], ["FastAPI"]]}
iex> Guided.Graph.query("""
MATCH (t:Technology {name: 'Streamlit'})-[:HAS_VULNERABILITY]->(v:Vulnerability)
RETURN v.name, v.severity
""")
{:ok, [["Cross-Site Scripting (XSS)", "high"], ["Path Traversal", "high"]]}We welcome contributors! To get started:
- Read the Development Setup Guide
- Check out GitHub Issue #1 for the full PRD
- Review docs/implementation_plan.md for the roadmap
- Look for issues labeled
good-first-issue
See CONTRIBUTING.md for detailed contribution guidelines.
- Fork the repository
- Create a feature branch (
git checkout -b feature/amazing-feature) - Make your changes and add tests
- Run
mix precommitto ensure quality - Commit your changes (
git commit -m 'Add amazing feature') - Push to your branch (
git push origin feature/amazing-feature) - Open a Pull Request
- Claude Desktop Setup - Quick start guide for Claude Desktop integration ⭐
- Development Setup - Complete setup guide
- MCP Server Guide - MCP server documentation and API reference
- AGENTS.md Specification - The protocol spec for AI agents
- Contributing Guidelines - How to contribute
- Implementation Plan - Phased roadmap
- Phase 0 Completion - Current progress
- Security Policy - Security guidelines
guided/
├── lib/
│ ├── guided/
│ │ ├── graph.ex # Graph database interface
│ │ └── repo.ex # Ecto repository
│ ├── guided_web/ # Phoenix web interface
│ └── mix/tasks/ # Custom mix tasks
├── priv/repo/ # Database migrations
├── config/ # Configuration files
├── docs/ # Documentation
└── scripts/ # Setup and utility scripts
- Phoenix Framework: https://www.phoenixframework.org/
- Apache AGE: https://age.apache.org/
- openCypher: https://opencypher.org/
- Model Context Protocol: https://modelcontextprotocol.io/
Apache 2.0
- Issues: GitHub Issues
- Discussions: GitHub Discussions
Built with ❤️ by the Guided.dev team