ertugrul37/sovereign-agent

> 🚀 1-Click Install in VS Code (vscode:extension/ertugrul37.sovereign-agent) · ⭐ Leave a 5-Star Review on Marketplace · ★ Star on GitHub
Sovereign Agent is an autonomous, privacy-first local AI coding agent for Visual Studio Code. Run Ollama, LM Studio, DeepSeek, Qwen, Gemma, Llama, llama.cpp, vLLM, LocalAI, Jan, or GPT4All directly inside your editor — with zero cloud calls, zero API keys, and zero telemetry.
## Why Sovereign Agent?
| Capability | Sovereign Agent | Cloud AI Assistants | Typical Local Plugins |
| Hard-Blocked Cloud & Telemetry (localGuard) | ✅ Enforced in code | ❌ Sends code to cloud | ⚠️ Optional / soft config |
| Autonomous File Read, Surgical Edit & Terminal | ✅ With approval gates | ✅ | ⚠️ Limited |
| Side-by-Side Model Arena (Compare 2 Local LLMs) | ✅ Built-in | ❌ | ❌ |
| LAN Peer-to-Peer Model Discovery & Sharing | ✅ Built-in (mDNS) | ❌ | ❌ |
| 1-Click Undo Checkpoint System | ✅ Built-in | ⚠️ Varies | ❌ |
| Local MCP (Model Context Protocol) Bridge | ✅ Loopback-guarded | ⚠️ Cloud/Mixed | ⚠️ Varies |
| Native UI Localization | ✅ 16 World Languages | ⚠️ Mostly English | ⚠️ English only |
## Key Features
### 🔒 1. Strict Local-Only Privacy (localGuard)
Every outgoing request passes through a strict loopback guard (127.0.0.1, localhost, ::1). Remote internet endpoints are hard-blocked at the network layer. Your proprietary source code never leaves your machine or local network.
### 🤖 2. Autonomous Agent with Explicit Approval Gates
Sovereign Agent can inspect your workspace, search code, perform surgical line replacements, create new files, and run terminal commands — always presenting a clear Allow / Deny confirmation card before touching your disk or shell, plus 1-click Undo Checkpoints.
### ⚔️ 3. Side-by-Side Local Model Arena
Not sure whether qwen2.5-coder, deepseek-coder-v2, or llama3.3 handles your codebase better? Select a Comparison Model to stream responses from two local models side-by-side in real time and pick the best answer.
### 🌐 4. LAN Peer Discovery & Team Model Sharing
Have a beefy GPU workstation on your local network? Enable allowLocalNetwork and networkShareEnabled to automatically discover and share local LLM endpoints across your LAN with zero external servers.
### 🌍 5. 16 Supported Languages
Switch the interface language anytime from the top header (🌐) or + menu:
English, Türkçe, Deutsch, Español, Français, 中文 (Simplified Chinese), Русский, 日本語, 한국어, Português, Italiano, العربية (RTL), Nederlands, Polski, हिन्दी, Azərbaycan.
## Supported Local AI Providers
| Provider | Default Local Endpoint |
| Ollama | http://localhost:11434 |
| LM Studio | http://localhost:1234/v1 |
| llama.cpp (llama-server) | http://localhost:8080/v1 |
| vLLM | http://localhost:8000/v1 |
| LocalAI | http://localhost:8080/v1 |
| Jan | http://localhost:1337/v1 |
| GPT4All | http://localhost:4891/v1 |
| Any OpenAI-Compatible Local Server | Custom localhost / LAN URL |