Andromity — AI Coding Agent for VS Code

The open-source coding agent that shows you exactly what it did. Waterfall, Fingerprint and Blast radius trace every model call, file read, edit and command. Bring your own API key or run local models with Ollama.

Website • Docs • Integrations • Quickstart • GitHub • Report Issue

Andromity is an open-source (MIT), BYOK (Bring Your Own Key) AI coding agent for VS Code. It plans multi-file tasks, edits code, runs commands and tests, and manages parallel subagents, while you review its plans, inspect native diffs, and roll back a turn with one click. Use it as an open-source alternative to the agent modes in Copilot or Cursor.

Connect your preferred AI provider (Anthropic Claude, OpenAI, Google Gemini, DeepSeek, Groq, OpenRouter, NVIDIA NIM) or run models locally and free with Ollama.

🔎 See Exactly What the Agent Did

Agents edit dozens of files and run commands faster than you can follow. Andromity records the whole run and gives you three views of it:

• 🌊 Waterfall. A live timeline of reasoning turns, model calls, terminal commands, file edits, web searches and parallel subagents. Inspect each span's timing and details. Click the 🌊 Waterfall button in the chat header, or run Andromity: Waterfall Trace.
• 🧬 Fingerprint. What the agent read, changed, ran and delegated, turn by turn: agent time, files changed, files read, commands run, and sensitive files accessed or blocked. Filter by changed, read, shell, subagents or sensitive; open any file at the line the agent touched; drill into a subagent's run; export the session as JSON.
• 💥 Blast radius. Each turn drawn as a hub with everything it touched around it, so you can see at a glance how far a change reached. Paths matching the agent's sensitive-path rules (.env files, SSH keys, cloud credentials, files named like secrets or tokens) are flagged.

Open Fingerprint and Blast radius from the Waterfall panel, or run Andromity: Fingerprint Trace & Blast Radius.

✨ Interactive Welcome Walkthrough

New to Andromity? The built-in walkthrough gets you oriented in about a minute:

1. Press Ctrl+Shift+P (Cmd+Shift+P on macOS) to open the Command Palette.
2. Select Andromity: Welcome / Getting Started (or use the welcome icon in the Andromity panel header).
3. Follow Get started with Andromity to tour the agent, planning, and workspace governance.

⚡ Quick Start

1. Install Extension

Click Install on this Marketplace page, or run code --install-extension agenticmarket.andromity-agent.

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

1. Open the Andromity panel in the Secondary Side Bar (press Ctrl+Alt+C / Cmd+Alt+C).
2. Click the ⚙️ Settings icon in the panel header.
3. Add the API key for your provider:
   • Anthropic (Claude)
   • OpenAI
   • Google Gemini
   • DeepSeek
   • Groq
   • OpenRouter
   • NVIDIA NIM
   • Ollama (free, local): no API key required; just have Ollama running on your machine. Pick a model that supports tool calling.

3. Trust Your Workspace and Start

Click Trust Folder in the chat panel (or type /trust) once you've reviewed the repository; the agent cannot read, edit, or run anything in an untrusted folder. Then ask a question or describe a feature you want to build.

🚀 How to Use Andromity in Your Workflow

💬 1. Interactive Sidebar Chat

• Chat with the agent about your workspace.
• Attach files or selected lines with Add to Andromity from the editor's right-click menu.
• Type @ to call a skill.
• The agent reads files, diagnoses bugs, implements features, and writes tests.

📝 2. Live Task Planner & Blueprints

For larger tasks, the Builder profile writes an implementation plan before editing.

• In SAFE and TRUST modes, review the plan, then approve it or send feedback. FULL and YOLO modes approve plans automatically.
• Watch each step's status update as the agent works.

🔍 3. Native Diff Review

• In SAFE mode, you approve each file edit before it is written.
• The Changes Review panel lists every file the agent changed, opens each one in a native VS Code diff, and lets you revert files individually.

⏪ 4. One-Click Rollback

Took a wrong turn? Run Undo Last Turn & Rollback Diff from the panel header or the Command Palette. Andromity restores the files changed in that turn without rewriting your Git history.

🤖 5. Multi-Agent & Parallel Subagents

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

🔌 6. App Connectors & MCP

Open Andromity Hub → MCP Servers → Discover (see every app and its setup guide) and connect Notion, Linear, Jira & Confluence, Gmail, Google Calendar, Google Drive, Stripe, Sentry, Supabase, Neon and more in one click. Each is the app's official remote MCP server; you sign in once in your browser and it works in every trusted workspace. Gmail, Calendar and Drive need your own Google OAuth client. GitHub, Airtable, Render and PagerDuty connect with an access token, which Andromity checks before saving.

Yours lists everything you've connected, with status, tools, re-authenticate and remove. You can still add any other MCP server through .andromity/mcp.json or .vscode/mcp.json. Tool schemas load on demand to keep token use down.

✏️ 7. Inline Edit

Select code and press Ctrl+I (Cmd+I), describe the change, then accept with Ctrl+Enter (Cmd+Enter) or reject with Escape.

🖱️ 8. Right-Click Context Menu

Highlight code, right-click, and open the Andromity menu:

• Inline Edit (Ctrl+I)
• Ask About Selection (Ctrl+Alt+A)
• Explain Code (Ctrl+Alt+E)
• Fix Diagnostics & Errors
• Generate Unit Tests (Ctrl+Alt+T)

🧩 9. Works with Copilot Chat

Mention @andromity in VS Code's Chat view to ask the Andromity agent to explain, fix, or build something.

✍️ 10. AI Git Commit Messages

Click the Andromity icon in the Source Control panel title bar to generate a commit message from your changes.

🌊 11. Waterfall, Fingerprint & Blast Radius

Trace every model call, file read, edit and command the agent made. See See Exactly What the Agent Did.

🔐 Trust Governance (Permission Modes)

Folder trust comes first: in an untrusted folder the agent cannot read files, edit them, or run commands, whatever the mode. Within a trusted folder, the permission mode decides what needs your approval:

Profile restrictions still apply in every mode: Planner and Reviewer never edit files or run commands. Switch modes anytime from the mode menu in the chat panel.

🤖 4 Specialized Agent Profiles

Switch profiles depending on what you're working on:

• 🏗️ Builder (default): Plans complex changes first, then implements them.
• ⚡ Coder: Implements directly, without a structured planning step.
• 🧐 Reviewer: Inspects code and reports findings without edits or commands.
• 📐 Planner: Creates plans without editing project code or running commands.

⌨️ 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 (default claude-sonnet-4-6).
• andromity.defaultProfile: Default agent profile (builder, coder, reviewer, planner).
• andromity.permissionMode: Default permission mode (safe, trust, full, yolo).
• andromity.reasoningEffort: Thinking effort for reasoning models (default auto; pick supported levels in the chat model picker, or budget:<tokens> for models with a thinking-token budget).
• andromity.soundNotifications: Turn completion sound cues on or off.
• andromity.telemetry: Anonymous usage and performance telemetry: launches, errors by category, and which features are used, by feature name only (on by default; also follows VS Code's telemetry setting).

🔒 Privacy & Security

• Where your code goes: to the model provider you choose. The built-in Andromity provider routes requests through the Andromity edge gateway, which forwards them without storing prompts, code, or completions.
• API keys: stored on your machine in the Andromity config file (~/.andromity/config.toml, or %APPDATA%\andromity\config.toml on Windows); keep that file private. Sign-in tokens use VS Code Secret Storage.
• Telemetry: anonymous usage and performance data: launches, session length, error categories, and which features are used (feature names such as fingerprint or cmd_openwaterfall). Never prompts, code, file paths, or model output. Turn it off with andromity.telemetry, VS Code's telemetry setting, or DO_NOT_TRACK=1.
• Local models: with Ollama, model inference runs on your machine. Web tools, remote MCP servers, and package installs can still use the network.

❓ FAQ

What is Andromity? An open-source AI coding agent for VS Code and the terminal. It reads your repository, proposes a plan, edits files, and runs commands such as your tests. Folder trust and permission modes control what it may do, and you choose the model.

Is it free? Yes. Andromity is open source under the MIT license. With your own API key, your provider bills its own usage; local Ollama models have no API charge.

Can it run locally with Ollama? Yes. Choose Ollama in the model picker and an installed model that supports tool calling. See the local Ollama guide.

Can the agent run commands without asking? Not in an untrusted folder. In a trusted folder, SAFE mode asks before each edit and command; FULL and YOLO run without prompts. See workspace trust and permission modes.

How does it compare with Claude Code, Cursor, Cline, or Aider? See the comparison table, checked against each tool's official documentation.

🌐 Links & Community

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