What license does OpenClaude use?

The README badge shows MIT, though the repository metadata lists the license as unknown, so check the LICENSE file in the repo to confirm before relying on it in a project.

Does OpenClaude require Claude Code to be installed?

No. OpenClaude stores its own config under `~/.openclaude` and does not read `~/.claude`, project `.claude/` directories, or Claude Code credentials, so new users don't need Claude Code installed.

Which LLM providers does OpenClaude support?

It supports OpenAI-compatible APIs (OpenRouter, DeepSeek, Groq, Mistral, LM Studio, and others), Gemini, GitHub Models, Codex (OAuth or CLI auth), Ollama for local models, and several gateway providers like Fireworks AI, LongCat, ApiSmart, Z.AI GLM, and Gitlawb Opengateway.

How do I run a fully local setup?

Point it at Ollama's OpenAI-compatible endpoint: set `CLAUDE_CODE_USE_OPENAI=1`, `OPENAI_BASE_URL=http://localhost:11434/v1`, and `OPENAI_MODEL` to a local model like `qwen2.5-coder:7b`, then run `openclaude`.

Can I run long-running agent tasks in the background?

Yes. `openclaude --bg " "` starts a detached local child process; use `openclaude ps`, `openclaude logs `, and `openclaude kill ` to manage it. There's no daemon or network service involved.

What are the system requirements?

Node.js `>=22.0.0` is required for npm installs and runtime; Bun is only needed for building from source. Ripgrep must be installed system-wide for search tooling to work.

OpenClaude: One terminal CLI to run coding

OpenClaude is an open-source, terminal-first coding-agent CLI for developers who want one consistent workflow across many LLM backends. Instead of learning separate tooling for each provider, you point OpenClaude at OpenAI-compatible APIs, Gemini, GitHub Models, Codex (OAuth or CLI auth), Ollama, Atomic Chat, and a growing list of gateway providers like Fireworks AI, LongCat, ApiSmart, and Z.AI GLM. It's aimed at developers who already use terminal coding agents (Claude Code-style workflows) but want provider flexibility, including running fully local models via Ollama.

Key features

Provider-agnostic CLI: switch between cloud APIs and local model backends without separate tooling per provider; guided setup via /provider saves profiles to .openclaude-profile.json.
Coding-agent toolset: bash execution, file tools, grep, glob, agents, tasks, MCP support, and web tools in one terminal session.
Session management: resume or continue conversations by session ID, fork history into new sessions with --fork-session, and inspect prior work.
Background sessions: run long prompts detached from the terminal with openclaude --bg, then track them with ps, logs, and kill subcommands (local child processes, no daemon).
Independent config: stores its own config under ~/.openclaude, doesn't read ~/.claude or Claude Code credentials, so it can run without Claude Code installed.
VS Code extension: bundled extension for launch integration and theming.
Repo map and smart routing: docs cover codebase intelligence, smart auto-routing, agent routing/step limits, and a headless gRPC server for automation.
Wide provider table: OpenAI-compatible endpoints (OpenRouter, DeepSeek, Groq, Mistral, LM Studio), Gemini, GitHub Models, Codex, Fireworks AI, LongCat, ClinePass, Gitlawb Opengateway, OpenCode Zen/Go, and others.

Ideal use cases

You want a single CLI workflow for coding agents but need to switch between multiple LLM vendors (cost, availability, capability) without rewriting integration code.
You run local models through Ollama for privacy or cost reasons and still want agent tooling (file edits, grep, bash, MCP) on top.
You maintain scripts or CI-adjacent workflows that need non-interactive, backgrounded agent runs (--bg, ps, logs) rather than a fully attached terminal session.
You're evaluating alternatives to a single-vendor coding CLI and want to test several model backends (OpenRouter, Fireworks, GitHub Models, Gemini) against the same task.
It's probably not a fit if you need a fully managed, GUI-based coding assistant, or if you want automatic loading of project .env files (OpenClaude intentionally doesn't do this by default). It's also not a good fit if you rely on full terminal reattach for background sessions today, since that feature isn't implemented yet, only log tailing is.

Installation

OpenClaude requires Node.js >=22.0.0 for npm installs and runtime. Bun is only needed if you're building from source.

Install globally via npm:

npm install -g @gitlawb/openclaude@latest

On Arch Linux, there's a community-maintained AUR package:

paru -S openclaude

If you see a ripgrep not found error, install ripgrep system-wide and confirm rg --version works before starting OpenClaude.

Verify or troubleshoot your installed version:

openclaude --version
npm view @gitlawb/openclaude dist-tags
npm install -g @gitlawb/openclaude@latest

Start the CLI:

openclaude

Inside the CLI, run /provider for guided provider setup, or /onboard-github for GitHub Models onboarding.

For a fast OpenAI-backed setup, set environment variables before launching:

export CLAUDE_CODE_USE_OPENAI=1
export OPENAI_API_KEY=sk-your-key-here
export OPENAI_MODEL=gpt-4o
openclaude

For a fully local setup with Ollama:

export CLAUDE_CODE_USE_OPENAI=1
export OPENAI_BASE_URL=http://localhost:11434/v1
export OPENAI_MODEL=qwen2.5-coder:7b
openclaude

Windows PowerShell equivalents are documented in the README for both setups. Additional beginner and advanced setup guides (Windows, macOS/Linux, non-technical, source builds) live under the docs/ directory in the repo.