index.md

LLxprt Code

LLxprt Code is an autonomous AI coding CLI that supports multi-pass development workflows. It works with any major LLM provider — Anthropic, OpenAI, Google Gemini, or local models — using your own subscriptions. No telemetry, no lock-in, fully open source under Apache 2.0.

Quick Start

npm install -g @vybestack/llxprt-code
llxprt

Once inside the REPL, authenticate with a provider and start coding:

/key save gemini your-gemini-api-key
/provider gemini
/model gemini-2.5-flash

Or use an API key with an open-weight model provider:

/key save synthetic syn****************b6
/provider Synthetic
/key load synthetic
/model hf:zai-org/GLM-4.7

For a full walkthrough including other providers, see the Getting Started Guide.

What You Can Do

Use Any Provider

Configure Claude, GPT, Gemini, or local models. Switch providers mid-session, set up multi-account failover, or load balance across API keys.

Recent open-weight models like DeepSeek, Kimi, Minimax, GLM, and Qwen are available through providers such as Z.ai, Synthetic, Chutes, Kimi.com, and Deepseek.ai.

• Provider Configuration
• Authentication
• Local Models

Multi-Pass Development

LLxprt Code plans, implements, tests, debugs, and iterates — across hours, not minutes. Delegate complex tasks to specialized subagents that run autonomously.

• Getting Started Guide
• Subagents

Your Subscriptions, Your Control

Use your existing Claude, OpenAI, or Google accounts via OAuth. Create profiles for different projects, teams, or workflows. Configure models, temperature, context limits, and more.

• Profiles
• OAuth Setup
• Configuration
• Prompt Configuration

Stay Safe

Run commands in sandboxed containers. Configure approval policies to control what the AI can do without asking. Store API keys and auth tokens in your system keyring — the LLM never has access to them. Use .llxprtignore to keep sensitive files out of context.

• Sandboxing

Extend It

Add capabilities through MCP servers, lifecycle hooks, custom themes, and more.

• Hooks
• MCP Server Integration
• Themes

Tools

LLxprt Code ships with built-in tools for file editing, shell commands, web search, code analysis, and memory. It also supports MCP servers and extensions for adding additional tools and capabilities. See the Tools Overview for the full list, including image generation.

Reference

• Agent API — embed the LLxprt Code agent in your own program
• Memory Profiling — record JSC heap samples and analyze heap snapshots to investigate session growth
• Application Directories — where config, data, cache, and logs live on Linux/macOS/Windows, and how to override them
• Token Usage Log — per-session token-usage JSONL schema, the join to conversation turns, and privacy posture
• Ephemeral Settings Reference — every ephemeral setting with defaults, types, and advice
• Profile File Reference — the profile JSON format, auth config, load balancers, precedence
• Commands
• Configuration
• Keyboard Shortcuts
• Troubleshooting
• Emoji Filter
• Context Dumping

Contributing

LLxprt Code is open source and community-driven. See the Contributing Guide to get started.