Andromity pairs natively inside your VS Code secondary sidebar or your terminal. It plans multi-file tasks, edits code, executes commands, and iterates until your tests pass, using your own API key or a local Ollama model.
Your files, conversation, and execution trace in one workspace. Keep your attention on the work.
Recording of the Andromity live execution waterfall in VS Code: model calls and file reads appear on a timeline as the agent worksYour workflow. Your model.
See how planning, execution, and review fit together inside your existing workspace.
Use the Builder profile to turn a task into an implementation plan. Review it, approve it, or send feedback before continuing.
Open the live waterfall to see tool calls, execution timings, and subagent activity. Inspect a span when you need to understand what happened.
Inspect additions and deletions in a native VS Code diff. If a turn takes the wrong direction, use Undo Last Turn to restore its file changes.
Install the agent, connect your model, and open a project. Start with a task you can review easily.
Run in your terminal, or install from the Marketplace.
code --install-extension agenticmarket.andromity-agentAndromity is 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: a cloud provider with your own API key or a local Ollama model.
Andromity is open source under the MIT license. With your own API key, your model provider sets the usage cost. You can also use Ollama locally without cloud API charges; performance depends on your hardware and model.
Andromity connects to models through LiteLLM, including Anthropic Claude, OpenAI, Google Gemini, Groq, OpenRouter, NVIDIA NIM, and OpenAI-compatible endpoints. It also runs local models through Ollama. Pick a model that supports tool calling; you can switch models during a session.
With a cloud model, the context needed for your task is sent to the provider you configure. The built-in Andromity provider routes requests through an edge gateway that forwards them without storing prompts, code, or completions. With local Ollama, inference runs on your machine. Folder trust and permission settings control what the agent can access and execute.
Not in an untrusted folder: there, file writes and commands are blocked whatever the permission mode. In a trusted folder, the permission mode decides how much needs approval, from SAFE, where you approve each plan, edit, and command, up to YOLO, which runs without prompts.
Yes. Andromity supports MCP servers so you can connect additional tools, plus skills and agent profiles to tailor the workflow to your project. MCP tool schemas load on demand instead of filling the prompt up front.
From punch cards to predictive text, developer tooling always evolved to elevate human leverage.
Terminal install requires pipx and Python 3.11+.