mtrinicli 2.0.0

Mtrini IDE 2.0 - Sovereign Local AI Coding Agent by Oryvia Labs

Mtrini Studio

Sovereign, local-first AI coding agent by Oryvia Labs. 100% local. Zero cloud. Zero subscriptions.

---

Free Tier

Mtrini Studio is free forever with 2 powerful AI models:

Model	Size	Best For
Mtrini Coder 2.0	~2.2 GB	Fast code completion, refactoring, debugging
Mtrini Life 2.0	~1.8 GB	General assistant, writing, explanations

No API key needed. No account required. No telemetry.

---

Quickstart

Install

pip install mtrinicli

Launch

mtrini              # Launch TUI (default)
mtrini --cli        # Launch CLI mode
mtrini --web        # Launch web server
mtrini --setup      # First-time setup wizard

Or run directly

python mtrini_ide.py
python mtrini_ide.py --cli
python mtrini_ide.py --setup

---

What You Get

• Local AI Agent - Reads, writes, and edits your code using local LLMs
• Plan & Build Modes - Read-only analysis or full file system access
• 8 Built-in Tools - Bash, read, write, edit, grep, glob, web search, web fetch
• BYOK Support - Use OpenAI, Anthropic, Google, Ollama, or any OpenAI-compatible API
• Desktop App - Native Electron window with dark glassmorphism UI
• MCP Integration - Connect external tool servers
• Cascade Fallback - Auto-downgrades models if your hardware can't handle the big ones

---

BYOK (Bring Your Own Key)

Use any LLM provider via the OpenAI-compatible API format:

Local Providers (Free, No API Key)

Provider	URL
Mtrini (Default)	localhost:8080
Ollama	localhost:11434
LM Studio	localhost:1234
llama.cpp	localhost:8080

Cloud Providers (API Key Required)

Provider	Models
OpenAI	GPT-4o, GPT-4o-mini, o1, o3-mini
Anthropic	Claude Sonnet 4, Claude Opus 4
Google Gemini	Gemini 2.5 Flash, Gemini 2.5 Pro
Groq	Llama 3.3 70B, Mixtral 8x7B
Together AI	Llama 3.3 70B, Qwen 2.5 Coder 32B
DeepSeek	DeepSeek Chat, DeepSeek Coder
OpenRouter	100+ models gateway

---

Architecture

mtrini-ide/
├── providers/          # BYOK provider system
│   ├── base.py         # BaseProvider abstract class
│   ├── openai_compat.py # Universal OpenAI-compatible adapter
│   ├── registry.py     # Provider registry + health checks
│   └── vault.py        # Encrypted API key storage
├── tools/              # Agent tool system
│   ├── base.py         # BaseTool with permission system
│   ├── registry.py     # Tool registry + execution
│   ├── bash.py         # Shell command execution
│   ├── read_file.py    # File reading with line numbers
│   ├── write_file.py   # File writing with auto-mkdir
│   ├── edit_file.py    # String replacement with diff preview
│   ├── grep.py         # Regex search (ripgrep + fallback)
│   ├── glob.py         # File pattern matching
│   ├── websearch.py    # DuckDuckGo web search
│   └── webfetch.py     # URL content fetching
├── engine/             # Agent loop engine
│   ├── agent.py        # Core agent loop (plan/build modes)
│   └── context.py      # Context window management
├── desktop/            # Electron desktop app
│   ├── main.js         # Electron main process
│   ├── preload.js      # Secure bridge
│   └── renderer/       # Frontend HTML/CSS/JS
├── app/                # FastAPI backend
│   ├── main.py         # OpenAI-compatible API server
│   ├── model.py        # llama-server process manager
│   └── mcp_manager.py  # MCP tool integration
├── models/             # Local GGUF model files
├── mtrini_ide.py       # Main entry point
└── config.yaml         # Configuration

---

Agent Modes

Plan Mode (Read-Only)

/plan    # Switch to plan mode

• Analyzes code without making changes
• Proposes implementation plans
• Identifies risks and edge cases

Build Mode (Full Access)

/build   # Switch to build mode (default)

• Reads, writes, and edits files
• Runs shell commands
• Full tool access with permission controls

---

Tool System

Tool	Permission	Description
bash	ask	Execute shell commands
read	allow	Read file contents
write	ask	Create/overwrite files
edit	ask	String replacement with diff
grep	allow	Regex content search
glob	allow	File pattern matching
websearch	allow	Web search via DuckDuckGo
webfetch	allow	Fetch URL content

---

Slash Commands

Command	Description
/help	Show all commands
/clear	Clear chat history
/plan	Switch to plan mode
/build	Switch to build mode
/provider NAME	Set active provider
/model NAME	Set active model
/stats	Show session statistics
/undo	Undo last response
/support	Support the project
/exit	Exit the IDE

---

Desktop App

cd desktop
npm install
npm start

Features:

• Native window with hidden title bar
• File tree sidebar
• Dark glassmorphism UI
• Settings panel for provider configuration
• System integration (menus, keyboard shortcuts)

---

Privacy & Security

• 100% Local: All inference runs on your hardware
• Encrypted Keys: API keys stored with XOR encryption in ~/.mtrini/vault.enc
• No Telemetry: Zero data sent to external servers
• Permission Controls: Granular tool permissions per session

---

Support the Project

If you enjoy Mtrini Studio, consider supporting its development:

• Ko-fi: ko-fi.com/ayh4m
• GitHub Sponsors: github.com/sponsors/oryvialabs

Your support helps keep Mtrini free and maintained.

---

About Oryvia Labs

Built by Oryvia Labs, founded by Ayham Habbad based in Salé, Morocco.

• PyPI: pypi.org/project/mtrini-ide
• License: Custom EULA - see LICENSE.txt