Litol AI
The Litol AI coding agent inside Visual Studio Code.
Connect your Litol AI gateway, choose your preferred model, and let the autonomous coding agent inspect, edit, run, and verify work directly inside your repository.
Litol AI brings a fast, transparent Codex-style agent loop to Visual Studio Code, backed directly by your Litol AI gateway. Open any project, choose a model, and describe the outcome you need. Litol AI inspects the codebase, plans the tasks, creates and edits files, executes commands, and displays every step in high contrast with real-time feedback.
Key Features
- Litol AI Gateway Native: Directly connects to your Litol AI gateway (
https://fedora.tail011e49.ts.net) with OpenAI-compatible SSE streaming, dynamic /v1/models discovery, and Bearer token auth.
- Dual Authentication: Sign in via one-click Browser OAuth or enter your Gateway API Key directly in Settings. Credentials persist securely in
~/.litol/gateway.json.
- Obsidian & Electric Cobalt Theme: Engineered on VS Code theme tokens with high-contrast obsidian backgrounds, Electric Cobalt (
#3b82f6) primary focus rings, and readable monospace font stacks.
- Autonomous & Guided Modes:
- Ask Mode: Proposes changes and opens diff editors for review before applying.
- Auto Edits Mode: Applies non-destructive file edits automatically; commands request approval.
- Open Access Mode: Full agent execution with automated verification.
- Project Intelligence: Fast, bounded local repository indexer cataloging exported symbols, imports, configuration files, and frameworks without transmitting your codebase to third parties.
- Git Safety & Checkpoints: Every turn records file modifications. Inspect per-file diffs, revert individual files, stage task changes, or roll back to pre-turn Git checkpoints with one click.
- Subagents & Tool Delegation: Delegates tasks to specialized subagents (explorer, reviewer, worker) with sandboxed execution contexts.
- Persistent Shell & Terminals: Keeps background shell processes alive across turns for dev servers, build watchers, and test runners.
- Extensible with MCP & Skills: Integrates Model Context Protocol (stdio, HTTP, SSE) servers and community
SKILL.md workflows.
Quick Start
1. Installation
Install the extension from the .vsix package:
code --install-extension litol-ai-0.1.0.vsix
2. Sign In to Your Gateway
- Open any workspace in VS Code.
- Click the Litol AI icon in the Activity Bar or press
Ctrl+Shift+P (macOS: Cmd+Shift+P) and run Litol AI: Open Chat.
- In the chat sidebar or in Litol AI: Open Settings, sign in:
- Direct API Key: Paste your Litol AI Gateway API key and endpoint (defaults to
https://fedora.tail011e49.ts.net).
- Browser Sign-in: Click Sign in with Browser to authenticate through the OAuth loop on
localhost:40822/callback.
- Your available models are discovered automatically from
/v1/models.
3. Start Coding
- Keep Auto selected to automatically select the most cost-effective eligible model, or choose any model from the dropdown.
- Type your prompt in the composer, or use slash commands like
/plan, /fix, /review, or /new.
Architecture & Gateway Integration
VS Code Extension Host (src/) ──► Webview UI (webview-ui/)
│
▼
~/.litol/gateway.json (Bearer Auth)
│
▼
Litol AI Gateway (https://fedora.tail011e49.ts.net)
├── /v1/models (Model Discovery & Pricing)
└── /v1/chat/completions (SSE Streaming & Tool Calls)
- Local Config: Credentials and active gateway settings are stored in
~/.litol/gateway.json.
- Worktrees: Git worktree sandboxes reside in
.litol/worktrees/.
- Plans: Multi-step implementation plans are saved in
.litol/plans/.
- Durable Memory: Workspace project memory is preserved in
.litol/memory.md.
Slash Commands
| Slash Command |
Action |
/plan <task> |
Create and maintain a pinned step-by-step implementation plan |
/fix <issue> |
Investigate an error, apply the fix, and run validation checks |
/review <scope> |
Perform code, safety, and regression reviews |
/test <scope> |
Run project test suites and report results |
/explain <topic> |
Explain architecture or logic without making changes |
/compact |
Compact current conversation history to preserve context window |
/context |
Open the active file, selection, and index context manager |
/new |
Start a fresh conversation |
/settings |
Open Litol AI settings |
/usage |
View token usage and gateway quota |
Configuration Settings
Settings can be managed through the VS Code Settings UI under litolAi.* or directly in settings.json:
| Setting |
Default |
Description |
litolAi.model |
"" |
Selected model ID for Litol AI gateway |
litolAi.maxSteps |
200 |
Maximum iterations per turn (0 for unlimited) |
litolAi.approvalMode |
"ask" |
Approval mode: ask, edits, or autonomous |
litolAi.repoIndex |
true |
Enable bounded local repository symbol indexing |
litolAi.worktrees |
true |
Enable isolated Git worktrees |
litolAi.systemNotifications |
true |
Native system notifications on completion or attention |
litolAi.mcpServers |
{} |
JSON map of configured Model Context Protocol servers |
Development
git clone https://github.com/litol-ai/litol-ai-extension.git
cd litol-ai-extension
npm install
npm run check
npm run build
To build a fresh .vsix package:
npm run package
License
MIT
| |