Andromity – Autonomous AI Coding Agent for VS Code

The only AI coding agent with trust governance, subagents, live plans, native diffs & one-click rollback.

Website • GitHub • Docs • Report Issue

Andromity is a private, BYOK (Bring Your Own Key) autonomous AI coding agent for VS Code. Designed as an extensible Copilot and Cursor alternative, it doesn't just autocomplete code — it plans out complex tasks with an interactive task planner, manages parallel subagents, shows you live step-by-step blueprints, lets you review diffs before applying, and gives you instant one-click rollback if you change your mind.

Connect your preferred AI model (Claude 3.7 Sonnet, GPT-4o, Gemini 2.5 Pro, DeepSeek R1 & V3, Groq, OpenRouter) or run 100% locally and free with Ollama.

✨ Interactive Welcome Walkthrough

New to Andromity? VS Code includes a built-in interactive walkthrough to get you oriented in under a minute:

1. Press Ctrl+Shift+P (Cmd+Shift+P on macOS) to open the Command Palette.
2. Select Andromity: Welcome / Getting Started (or click the 🚀 rocket icon in the Andromity panel header).
3. Follow the guided walkthrough to tour your AI coding partner, inspect planning blueprints, and adjust workspace governance.

⚡ Quick Start — Up and Running in Seconds

1. Install Extension

Click Install on this Marketplace page. That's it!

2. Add Your API Key (or Use Local Ollama)

1. Open the Andromity tab in your Activity Bar (left sidebar).
2. Click the ⚙️ Settings icon in the header.
3. Paste the API key for your favorite provider:
   • Anthropic (Claude 3.7 / 3.5 Sonnet)
   • OpenAI (GPT-4o, o3-mini)
   • Google Gemini (Gemini 2.5 Pro / Flash)
   • Groq (Llama 3.3 70B, DeepSeek)
   • DeepSeek / OpenRouter
   • Ollama (Free / Local): No API key required! Just have Ollama running locally on your machine.

3. Start Coding!

Press Ctrl+Alt+C (Cmd+Alt+C on macOS) or open the sidebar. Ask a question, select a file, or describe a feature you want to build.

🚀 How to Use Andromity in Your Workflow

💬 1. Interactive Sidebar Chat

• Chat directly with the AI about your workspace.
• Reference files with @ mentions.
• The agent reads files, diagnoses bugs, implements features, and writes tests.

📝 2. Live Task Planner & Blueprints

When tackling complex tasks, Andromity creates an interactive implementation plan first.

• Review steps before changes happen.
• Approve, skip, or edit steps individually.
• Watch real-time execution status for each task.

🔍 3. Side-by-Side Diff Review

Every change proposed by the agent is presented as a native VS Code diff.

• See exact additions and deletions.
• Accept changes file-by-file or accept all at once.

⏪ 4. One-Click Rollback (/undo)

Made a turn you don't like? Simply trigger Undo Last Turn from the header menu or command palette. Andromity rolls back all file changes from that turn instantly without touching your git history.

🤖 5. Multi-Agent & Parallel Subagents

Spawn background subagents for research, file scanning, or isolated task implementation while you continue chatting in the main session.

🔌 6. Model Context Protocol (MCP) Ready

Connect any MCP server (tools, documentation, database connectors) directly into the agent workflow. Schemas load on demand to minimize token usage.

🖱️ 7. Right-Click Context Menu Superpowers

Highlight any snippet in your code editor, right-click, and choose:

• Ask About Selection (Ctrl+Alt+A): Inquire about logic or behavior.
• Explain Code (Ctrl+Alt+E): Get a breakdown of what the code does.
• Fix Diagnostics & Errors: Auto-resolve type errors, linter warnings, or crashes.
• Generate Unit Tests (Ctrl+Alt+T): Auto-generate complete test suites for your functions or classes.

✍️ 8. AI Git Commit Messages

Click the Andromity icon directly in the Source Control panel title bar to generate a clean, descriptive commit message based on your staged changes.

🔐 Trust Governance (Permission Modes)

You are always in control of what the agent does in your workspace:

Switch modes anytime by clicking the permission badge in the top bar or via Andromity: Switch Permission Mode.

🤖 4 Specialized Agent Profiles

Switch profiles depending on what you're working on:

• 🏗️ Builder (default): Plans out tasks first, then implements them.
• ⚡ Coder: Fast direct implementation without an upfront planning stage.
• 🧐 Reviewer: Read-only mode for code audits, security checks, and PR reviews.
• 📐 Planner: Generates architectural specs and technical plans without modifying code.

⌨️ Keyboard Shortcuts

⚙️ Extension Settings

Customize Andromity in Settings (Ctrl+, → search Andromity):

• andromity.defaultProvider: Default AI provider (anthropic, openai, google, groq, openrouter, ollama, nvidia).
• andromity.defaultModel: Default model (e.g., claude-sonnet-4-6, gpt-4o, gemini-2.5-pro).
• andromity.defaultProfile: Default agent profile (builder, coder, reviewer, planner).
• andromity.permissionMode: Default trust mode (safe, trust, full, yolo).
• andromity.reasoningEffort: Thinking level for reasoning models (off, low, medium, high).
• andromity.soundNotifications: Enable or disable turn completion sound cues.

🔒 Privacy & Security First

• Zero Data Collection: Your code and prompts go strictly between your machine and your chosen AI provider.
• Local Secret Storage: API keys are encrypted locally using VS Code's native Secret Storage API.
• 100% Offline Capability: Use Ollama to keep all code and conversations strictly local on your machine.

🌐 Links & Community

• 🏠 Website: andromity.agenticmarket.dev
• 🐙 GitHub: agenticmarket/andromity
• 📋 Changelog: CHANGELOG.md
• 💬 Discussions & Feedback: GitHub Discussions
• 🐛 Bug Reports & Requests: GitHub Issues