Litos for VS Code

Your open-source AI coding agent inside VS Code.

Litos understands your codebase, edits files, runs commands, fixes errors, and completes multi-step development tasks without making you leave your editor.

No Docker, database, or Litos account required.

Litos adding a copy feature to a PostgreSQL query app

What Litos can do

Understand and work across your open workspace
Read, create, and modify files
Run builds, tests, scripts, and shell commands
Work through multi-step development tasks
Use Claude, GPT, Gemini, OpenRouter, or local models
Connect to user-registered MCP servers
Attach files and paste images into a session
Maintain multiple independent chat sessions
Resume sessions across Litos clients
Reflect session learnings into AGENTS.md through a reviewable diff

Cloud or local models

Use cloud models from Anthropic, OpenAI, Gemini, and OpenRouter — or connect Litos to local models running through Ollama, LM Studio, and compatible endpoints.

Getting started

Install Litos from the Visual Studio Marketplace.
Open the Litos icon in the VS Code activity bar.
Run /keys and set up a provider: add an API key for a cloud provider (Anthropic, OpenAI, Gemini, or OpenRouter), or a local server URL for a local model (e.g. http://localhost:1234/v1 for LM Studio — most local servers don't need a key). This is stored once and shared with any other Litos client on your machine (CLI, desktop app).
Run /provider to choose a cloud provider or local endpoint, then /model to pick a model.
Open a project and describe the change you want.

No Docker, database, or Litos account is required for the VS Code extension.

Slash commands

Type / in the chat box to see the full list. These match the commands in the Litos desktop app, so a session works the same way in either client.

Command	What it does
`/keys`	Add or update your API key for a model provider
`/provider`	Switch chat provider (Anthropic, OpenAI, Gemini, OpenRouter, or a local endpoint)
`/model`	Switch model within the current provider
`/new`	Start a new session
`/resume`	Resume a previous session
`/skills`	List available skills
`/skill`	Load a skill by name
`/attach`	Attach a file
`/branch`	Branch a new session from an earlier message
`/compact`	Compact the conversation now to free up context
`/reflect`	Distill this session into `AGENTS.md`, opened in a reviewable diff
`/mcp`	Manage MCP servers

MCP support

Connect, manage, and use Model Context Protocol servers from a dedicated panel. Litos's built-in coding tools are fixed, minimal, and auto-approved; MCP servers you register are the integrations you choose to enable or disable.

Sessions are shared across Litos clients

Litos stores your conversation history locally under ~/.litos/sessions, the same place used by the Litos CLI and desktop app. A session started in one client can be resumed from another.

Supported operating systems

First-class support: Windows and macOS, including Intel and Apple Silicon Macs. Linux support is planned but is not currently available.

Local by design

The bundled Litos process listens only on 127.0.0.1. It is not exposed to other computers on your network. Your prompts and relevant code are sent only to the model provider or local endpoint you configure.

Sessions are stored locally under ~/.litos/sessions.

Pricing

Litos is free and open source. Bring your own provider API key, or use a compatible model running locally on your machine. Cloud-provider usage charges may apply.

Litos

Litos AI