Lightweight terminal AI coding agent — your AI pair programmer in the terminal.
pip install bonza-mragent
mr
- Multi-Provider: Ollama, NVIDIA NIM, OpenRouter, OpenAI, Anthropic, xAI/Grok, Qwen
- Arrow-Key TUI: Navigate providers and models with arrow keys — no typing numbers
- Agentic Tools: File editing, shell execution, grep/search, git integration
- Web Browsing: Browse the web without search APIs (optional Playwright)
- Chat History: SQLite-backed sessions — resume, search, export
- Cron Scheduling: Schedule recurring agent tasks
- Smart Routing: Architect/Editor pattern for complex tasks
- Token Tracking: Real-time token count, context usage, and cost display
- Lightweight: ~13MB core install, pure Python
# Install
pip install bonza-mragent
# Run (first time shows interactive setup)
mr
# Or specify provider directly
mr --provider ollama --model llama3.1:8b
# Setup / reconfigure
mr setup
# Browse chat history
mr history| Provider | Setup | Free Tier |
|---|---|---|
| Ollama | ollama serve locally |
✅ Local |
| NVIDIA NIM | API key from build.nvidia.com | ✅ Credits |
| OpenRouter | API key from openrouter.ai | ✅ Limited |
| OpenAI | API key from platform.openai.com | ❌ |
| Anthropic | API key from console.anthropic.com | ❌ |
| xAI / Grok | API key from console.x.ai | ❌ |
| Qwen | API key from dashscope.console.aliyun.com | ✅ Trial |
| Command | Description |
|---|---|
/provider |
Switch AI provider |
/model |
Switch model |
/history |
Browse chat history |
/resume <id> |
Resume old chat |
/cost |
Show token usage & cost |
/cron |
Manage scheduled tasks |
/export |
Export chat as Markdown |
/compact |
Compress context window |
/help |
Show all commands |
pip install bonza-mragent[browser]MIT