vybestack/llxprt-code
An open-source multi-provider AI assisted CLI development tool. Use whatever LLM you want to code in your terminal.
What it solves
Developers often get locked into a single AI assistant like GitHub Copilot or ChatGPT, or have to use separate tools for different providers. LLxprt Code solves that by giving you one command-line tool that works with any LLM provider — Anthropic, Gemini, OpenAI, Kimi, or any OpenAI-compatible service — and can even run completely offline with local models. It lets you use existing subscriptions (Claude Pro/Max, ChatGPT Plus/Pro, Kimi) via OAuth, or free tiers like Gemini via a Google account, so you don't need to pay extra for a coding assistant.
How it works
It's a CLI you run in your terminal. You start with llxprt, then pick a provider with /provider and choose a model with /model. You can ask questions, generate code, analyze a codebase, or debug errors in an interactive REPL, or run single commands in non-interactive mode for automation and CI/CD. Under the hood, it uses the Bun runtime (bundled with the package) so you don't need to install anything else. It supports multiple accounts and load balancing across providers for automatic failover when you hit rate limits. You can create subagents — isolated assistants with different models, providers, and tool access — and connect to external tools via MCP servers.
Who it’s for
Developers who want a flexible, privacy-conscious AI coding assistant that isn't tied to one vendor. It's especially useful for people who already have a Claude, ChatGPT, or Kimi subscription and want to use it in the terminal, or for those who want to run models locally with LM Studio, llama.cpp, or Ollama. It also fits teams that need to automate coding tasks in scripts or CI/CD pipelines.
Highlights
- Works with any LLM provider: Anthropic, Gemini, OpenAI, Kimi, or any OpenAI-compatible API, plus local models via LM Studio, llama.cpp, and Ollama.
- Use existing subscriptions (Claude Pro/Max, ChatGPT Plus/Pro via Codex, Kimi) through OAuth, or free tiers like Gemini with a Google account.
- Multi-account failover and load balancer profiles automatically route requests to available providers or accounts.
- Subagents: create isolated assistants with different models, providers, and tool access for specialized tasks.
- MCP integration for connecting to external tools and services.
- No telemetry by default — your data stays local unless you choose to send it to an external provider.
- Interactive REPL with a beautiful terminal UI and multiple themes, plus non-interactive mode for scripts and CI.
- Native Zed editor integration via the Agent Communication Protocol (ACP).
Related
- Project
- Project
- Project
- Project