OpenAI-Compatible Local CLI Coding Agent

Standalone AI coding agent for local LLM environments. Works with vLLM, Ollama, LM Studio, and any OpenAI-compatible API.

# 1. Install
git clone https://github.com/A2G-Dev-Space/Local-CLI.git
cd Local-CLI
npm install && npm run build

# 2. Run
node dist/cli.js       # or use 'lcli' command after npm link

The LLM endpoint setup wizard will automatically run on first launch.

Request user approval before executing file modification tools.

┌─────────────────────────────────────────────────────────────┐
│  🔧 create_file                                              │
│  ─────────────────────────────────────────────────────────   │
│  📁 file_path: /src/utils/helper.ts                          │
│  📝 content: export function helper() { ... }                │
│  ─────────────────────────────────────────────────────────   │
│  ▸ [1] ✅ Approve                                            │
│    [2] ❌ Reject                                             │
└─────────────────────────────────────────────────────────────┘

• Tab key - Toggle Auto ↔ Supervised mode
• Only file modification tools require approval (read_file, list_files, etc. run automatically)
• On Reject - Enter comment → AI retries with feedback

Automatically breaks down user requests into TODO lists and executes them sequentially.

You: Add a logging system to the project

✶ Thinking… (esc to interrupt · 5s · ↑ 1.2k tokens)

┌────────────────────────────────────────────────┐
│ 📋 TODO List                            1/3    │
│ ████████░░░░░░░░░░░░░░░░░ 33%                  │
│ ├─ ☑ Create logger.ts file                    │
│ ├─ ⣾ Add logger import to existing files      │
│ └─ ☐ Apply logger to error handling           │
└────────────────────────────────────────────────┘

✶ 1/3 tasks · edit_file… (esc to interrupt · 12s)

Claude Code-style scrollable log history:

• Tool-specific icons (📖 read, 📝 create, ✏️ edit, 📂 list, 🔍 find, 🔧 bash, 💬 message)
• Diff format for file changes (blue: added, red: deleted)
• Real-time progress display

• Ctrl+C - Exit
• ESC - Interrupt current execution
• Tab - Toggle Auto ↔ Supervised mode
• @ - File browser
• / - Command autocomplete

# --eval mode usage
echo '{"prompt": "1+1은?"}' | lcli --eval

# Run Python tests
npm run test        # Full test
npm run test:quick  # Quick test

• Request user approval before file modification
• Toggle Auto/Supervised mode with Tab key
• Provide feedback via comments on Reject

• Auto-save/restore conversation history between TODO tasks
• Preserve full context including tool calls/responses
• History only resets on /compact

• Status bar shows Context (1.3K / 13%) format
• Auto-Compact runs automatically at 80% usage

• parallel_tool_calls: false API parameter enforced
• LLM calls only one tool at a time for stable execution

# Run setup wizard
lcli    # First run auto-launches wizard

# Or via settings
/settings

Compatible with any OpenAI-compatible API server:

• vLLM, Ollama, LM Studio
• Azure OpenAI, Google Gemini (OpenAI Compatible)
• Internal LLM servers

lcli              # Default run
lcli --verbose    # Verbose logging
lcli --debug      # Debug mode

~/.local-cli/
├── config.json        # Configuration file
├── endpoints.json     # Endpoint settings
├── usage.json         # Usage statistics
├── docs/              # Downloaded docs
└── projects/          # Project-specific sessions

• Node.js v20+
• npm v10+
• Git (for doc downloads)

• Developer Guide
• Logging System
• Testing Guide
• Roadmap

MIT License

AI coding assistant local LLM offline AI CLI tool vLLM Ollama LM Studio OpenAI compatible code generation developer tools TypeScript Node.js coding agent

GitHub: https://github.com/A2G-Dev-Space/Local-CLI