Twigpine/openclaude

OpenClaude – an open‑source coding‑agent CLI

What it is – OpenClaude is a terminal‑first command‑line interface that lets you talk to a wide range of large‑language‑model back‑ends (OpenAI, Gemini, Ollama, GitHub Models, many commercial gateways, etc.) and use the model as a coding assistant. It bundles a set of tools (bash execution, file read/write, grep/glob, web search, custom agents) and a simple session system so you can keep, resume, or branch conversations.

Key features

• One CLI for many providers – configure any OpenAI‑compatible API, Gemini, Ollama, or the listed partner gateways via the /provider command or environment variables.
• Coding‑agent workflow – built‑in tools for running shell commands, editing files, searching the codebase, and invoking custom agents or “skills”.
• Session management – save, list, resume, delete, or fork chat sessions; background sessions can run non‑interactive prompts detached from the terminal.
• VS Code extension – launch OpenClaude from the editor and get theme support.
• Cross‑platform – works on macOS, Linux, Windows (Node ≥ 22). An AUR package is available for Arch Linux.
• Extensible – supports custom slash commands, smart auto‑routing, gRPC server mode, repo‑map intelligence, and Android installation.

Installation

# Global npm install (Node >= 22)
npm install -g @twigpine/openclaude@latest

If you previously installed the older @gitlawb/openclaude package, uninstall it first. On Arch Linux you can also use paru -S openclaude. After install, run openclaude to start the REPL.

Typical workflow

1. Configure a provider – inside the REPL run /provider (or set env vars) to select a cloud API key, model, and endpoint.
2. Start a session – just type a prompt, e.g. Write a function that validates an email address.
3. Use tools – slash commands let you run bash, edit files, grep, or call web search (/search).
4. Manage sessions – /sessions lists saved chats; openclaude --resume <id> or --continue resumes them; --fork-session creates a new branch of the conversation.
5. Background jobs – openclaude --bg "fix failing tests" runs a prompt in the background; openclaude ps, logs, and kill manage those jobs.

Supported back‑ends (selected) – OpenAI, OpenRouter, DeepSeek, Mistral, Gemini, Ollama (local), GitHub Models, Codex, Fireworks AI, LongCat, Cloudflare Workers AI, Bedrock/Vertex, plus many partner gateways (AI/ML API, Novita, ApiSmart, Concentrate, Exa, GMI Cloud, Xiaomi MiMo, etc.). The CLI automatically discovers available models from each provider.

Community & ecosystem

• Discussions & Discord – active channels for help, feature ideas, and troubleshooting.
• VS Code extension – launch OpenClaude from the editor and get a pixel‑art “hero” companion.
• Extensive docs – quick‑start guides for non‑technical users, Windows/macOS/Linux specifics, and advanced topics (smart routing, repo‑map, gRPC server, Android install).
• Open‑source – MIT‑licensed, published on npm and mirrored to a public Git repository.

OpenClaude is essentially a unified front‑end for LLM‑powered coding assistance, letting developers stay in the terminal while swapping between cloud and local models without juggling separate SDKs or CLIs.