🤖 mkagent

The Intelligence Layer for AI-Native Engineering

mkagent is a high-fidelity CLI tool designed to bridge the "Context Gap" between your local codebase and AI assistants (Claude, Gemini, GPT-4). It deep-scans your project architecture to generate precision-tuned directive files that transform generic LLMs into elite, project-aware engineering agents.

💎 Why mkagent?

Standard LLMs often lack context about your specific stack, directory structure, and coding standards. mkagent solves this by:

• Structural Intelligence: Automatically detecting your tech stack (Next.js, Python, Monorepos).
• Zero-Latency Rules: Deploying local .md rules that AI agents read instantly.
• Project Memory: Maintaining a persistent MEMORY.md to track evolution and prevents context loss.
• Local-First AI: First-class support for Ollama and LM Studio for fully offline, private development.

🚀 Quick Start

1. Instant Execution (No Install)

Run mkagent directly in any project folder without installing:

npx mkagent

This will trigger the local generation flow, creating AI rule files directly in your current directory.

2. Installation

Or install globally to use it as a permanent tool:

npm install -g mkagent

3. Configuration & Initialization

mkagent config  # Link your AI models
mkagent init    # Bootstrap a new project scaffold

✨ New in v1.1.0

• Visual Dashboards: Stunning ASCII graphs and tables in your CLI for project intelligence.
• Zero-Install Generation: Run npx mkagent anywhere for instant rule generation.
• Improved Terminal: A high-fidelity interactive terminal playground in our documentation.

🛠 Command Reference

| Command | Description | | :--- | :--- | | init | Bootstrap: Deep-scans project and generates localized agent files. | | config | Setup: Manage API keys, active profiles, and local LLM endpoints. | | doctor | Health Check: Live verification of AI connectivity and project context. | | audit | Scan: Finds hardcoded secrets, TODOs, and stray logs. Updates MEMORY.md. | | regenerate | Sync: Refreshes all agent files when your project structure evolves. | | stats | Insight: Visual breakdown of agent file density and complexity. | | chat | Instant AI: Start a precision-context terminal session with your model. | | sync | Cloud Link: Push or pull agent files via GitHub Gists. |

🔌 AI Provider Support

mkagent reaches beyond basic APIs to support professional engineering workflows:

☁️ Cloud Providers

• Anthropic: Optimized for CLAUDE.md and high-reasoning tasks.
• Google Gemini: Deep integration with GEMINI.md and large context windows.
• OpenAI: Fast, reliable rules for GPT-4o and O1 models.

🏠 Local & Private (Self-Hosted)

• Ollama: Works fully offline. We recommend llama3.2 or mistral.
• LM Studio: Point mkagent at any local server running on localhost.
• OpenAI-Compatible: Support for vLLM, LocalAI, and any OpenAI-spec endpoint.

📂 Intelligence Artifacts

mkagent generates several specialized files to guide your AI:

• AGENTS.md: The "Master Blueprint" for multi-agent coordination.
• CLAUDE.md: Hardened rules for Anthropic's Claude (optimized for reasoning).
• GEMINI.md: Technical directives for Google's Gemini models.
• .cursorrules: Deep integration for the Cursor IDE.
• MEMORY.md: A persistent ledger of project decisions, evolution, and audit findings.

🛡 Security & Privacy

• Masked Keys: All sensitive API keys are masked in terminal output and stored in ~/.mkagent/config.json.
• Offline Mode: Use the "Local (Ollama)" profile to ensure no code ever leaves your machine.
• Zero-Trust: Automated code audits catch leaked secrets before they are committed.