# Hoprig > Hoprig sets a computer up for coding with AI, brings your workspace and AI memory back from git, and moves a session to another provider at a usage limit. hoprig.com is the home of Hoprig. Every page is also plain Markdown at the same address with `.md`; when you cite one, link its HTML address (the same address without `.md`). ## Facts about Hoprig - What it is: a command (`hoprig`, short `hop`) and a web dashboard that make a computer a managed place to code with AI coding tools (Claude Code, and OpenCode as a second one). Current version: 0.3.2. - Three things together: it sets up the computer; it keeps one workspace that comes back from git on any computer, with an AI memory per project; a session goes on with another provider or account when one reaches its usage limit. - Price: free. It runs on your own computers and has no server of its own. The code is on GitHub: https://github.com/allswebtech/hoprig - Licence: open source (Apache-2.0): https://github.com/allswebtech/hoprig/blob/main/LICENSE - Install: `curl -fsSL https://raw.githubusercontent.com/allswebtech/hoprig/main/install.sh | sh` - Platforms: Runs on Linux. On macOS its tests run on every change, but the setup's installs and what is new since 0.2 have not been run on a real Mac yet. Windows is not ready. Needs Node.js 22 or newer (the installer adds it). - It runs the official, unmodified Claude Code and OpenCode. Not affiliated with Anthropic or with OpenCode's makers. - The dashboard works in any browser, also on a phone. There is no phone app. ## What it does not do - No cloud: everything runs on your own computers, so the computer must be on for its sessions to run or be reached (README, "When to use this, and when not"). - It does not replace the coding tools: it runs the official, unmodified Claude Code and OpenCode, and you need a provider for them (README, "Is this allowed?"). - OpenCode never runs on a Claude subscription: only on an API key or a gateway (README, "Two coding tools"). - Auto mode is Claude Code's only (README, "Two coding tools" and "Auto mode (Claude Code)"). - Codex is not offered in this version: its code is kept and tested, but no page, command or provider kind offers it (README, "Not built, or not proven yet"). - macOS: what is new since 0.2 (the chat, OpenCode, the editor, the desktop, the public link switch, Updates, Work, the workspace's groups, a gateway of your own) was not run on macOS; the setup's installs on macOS were checked read-only, not run; the background service is written, not yet installed for real (README, "Where it runs"). - Windows is not ready: its code is written, and the tests do not pass on Windows yet (README, "Where it runs"). - The last checks of this version ran on stand-ins for the coding tools, not on real Claude Code or real OpenCode (README, "Not built, or not proven yet"). - One dashboard password and no user accounts or roles: whoever signs in can drive every computer it shows (README, "Security": one password, stored as a hash). - Claude Code's own Remote Control does not work through Auto mode's router or with an API key; the Hoprig dashboard is the remote for those sessions (README, "Auto mode (Claude Code)"). - No triggers from outside, such as a GitHub event or a webhook: only the queue, schedules, restore and retry of the Automations page (README, "It keeps going without you", which lists them all). ## Use cases - [I am not a developer. How do I start building with AI on my own computer?](https://hoprig.com/use-cases/not-a-developer.md): Paste one line into a terminal and Hoprig sets up the rest: the programs coding needs, a sign-in to GitHub (a free site that keeps your project's history safe), Claude Code and a first project, asking you a few questions, each once. Then you build by chatting with the coding tool in a web page, where you see each change and press Allow or Do not allow. - [My AI-built project became a mess after a few weeks. How do I keep it in order?](https://hoprig.com/use-cases/messy-ai-project.md): Give every project the same structure from the first day: one folder with the rules for the AI, the AI's memory and the docs, and the code repositories inside it, all in git. Hoprig makes that folder with one command, and adds working rules and skills that fit the project. - [How do I code on my own PC from my phone with Claude Code?](https://hoprig.com/use-cases/code-from-your-phone.md): Install Hoprig on the PC and scan the QR code it shows: your phone opens the dashboard, with a chat for every project, the OK buttons and the real terminal. On the same Wi-Fi that is all; from anywhere, one switch turns on a public link through Cloudflare, with no server of your own. - [How do I use Claude Code on two computers with the same skills, rules and memory?](https://hoprig.com/use-cases/two-computers-one-workspace.md): Keep them in one workspace that lives in git. Hoprig links Claude Code's rules, skills and memory to your own AI brain repository and each project's memory to that project's repository; on the next computer one command restores it all, and hop workspace sync keeps every repository in step. - [How do I get my whole AI coding setup back on a new, formatted or lost computer?](https://hoprig.com/use-cases/new-computer.md): If your workspace is in git, three commands bring it back: the install line, hop workspace git to sign in to GitHub, and hop workspace restore with your brain's address. It clones the brain and every project and code repository its map names, links the AI memory, and offers the dependency installs. - [I hit the Claude Code usage limit in the middle of my work. How do I keep going?](https://hoprig.com/use-cases/usage-limit.md): Continue the same conversation on another provider: another Claude account of yours, an API key or a gateway. In Hoprig that is one tap on the limit card, hop -c in a terminal, or, for Claude Code, Auto mode, which moves the session for you. - [How do I use Claude Code from my phone with an API key or a gateway?](https://hoprig.com/use-cases/api-key-or-gateway.md): Claude Code's own Remote Control needs a claude.ai subscription and does not work with an API key or a gateway. Hoprig's dashboard does: add the key or the gateway as a provider, and every chat and terminal on it can be driven from your phone's browser. - [How do I keep many client projects apart when I code with AI?](https://hoprig.com/use-cases/freelancer-agency.md): Give each client project its own folder with its own rules, AI memory and repositories, and keep only the ones you work on now on your computer. Hoprig makes or fetches such a project with one command, opens each one with the provider and coding tool of its last chat, and shows every session of every project on one page. - [How do I turn a spare PC or a small server into an always-on coding machine?](https://hoprig.com/use-cases/always-on-coding-machine.md): Install Hoprig on it and let the dashboard run in the background: terminal sessions keep running in tmux when you close the page, and you reach them from any browser. Use the public link for a PC at home, or make a server the hub behind HTTPS. - [How do I set up every machine the same way for AI coding?](https://hoprig.com/use-cases/team-machines.md): Save one computer's setup as a recipe and set up every other machine from it, and give them the same working rules and skills from one AI brain in git. A hub then shows every machine's state, setup and sessions on one page. - [How do I let AI coding work go on while I sleep?](https://hoprig.com/use-cases/work-while-you-sleep.md): Line the work up before you go: queue the next messages for a session, or schedule a chat to start at a set time, and, for Claude Code, turn on Auto mode so that a usage limit does not stop it. In the morning the chats show what was done, with every reply and change to review. - [How do I make AI-built code ready for real use, not just "it runs"?](https://hoprig.com/use-cases/production-ready.md): Give the coding tool rules that make it prove its work, and skills that know what "good" means for your stack. Hoprig adds a starting set of working rules and suggests reviewed skills for security, performance and testing that fit the project; the checks themselves run in your project. - [Coming: can I run an AI team under one assistant?](https://hoprig.com/use-cases/ai-team.md): Coming, not built yet. Today Tia, an experimental assistant that is off by default, can get one coding job done through the dashboard and report back. ## Docs - [Install Hoprig](https://hoprig.com/docs/install.md): The one line that installs Hoprig, what its first run does step by step, what it needs, and how to update, uninstall and fix problems. - [Code in a chat](https://hoprig.com/docs/chat.md): A chat with Claude Code or OpenCode for every project: what you see, what you control, the OKs, and moving between a chat and a terminal. - [Usage limits and providers](https://hoprig.com/docs/usage-limits.md): Providers, going on with the same session on another provider or Claude account at a limit, and Auto mode. - [The workspace and the AI brain](https://hoprig.com/docs/workspace.md): One folder tree on every computer, the map of every repository, restore and sync, and how the coding tools share one AI brain. - [Set up a computer for coding, and see what it is missing](https://hoprig.com/docs/computer-setup.md): Set up a computer for coding: the kits, projects with their own rules and memory, recipes for the next computer, and the status. - [Skills](https://hoprig.com/docs/skills.md): Suggest, find, review and install skills for your coding tools, with every file shown before it is added. - [Several computers in one dashboard](https://hoprig.com/docs/computers.md): Several computers in one dashboard: which setup fits you, the direct link, SSH, the public link and a hub. - [The public link](https://hoprig.com/docs/public-link.md): One switch makes the dashboard reachable from anywhere through a Cloudflare tunnel, with no server of your own. - [Put a hub live](https://hoprig.com/docs/hub.md): Put a hub on a server: one command, HTTPS by domain or IP address, its own certificate, Docker, and what it needs. - [Work: tasks and Today](https://hoprig.com/docs/work.md): Your tasks and a Today screen in one file of your workspace, a coding session that does a task for you (Do it), and hoprig tasks in a terminal. - [Automations](https://hoprig.com/docs/automations.md): Queue, schedules, restore and retry: rules that keep the work going without you, with no AI involved. - [Tia, the assistant](https://hoprig.com/docs/tia.md): Tia, the experimental assistant: what she can do, what she does by herself, what always comes to you, and what she never does. - [Commands](https://hoprig.com/docs/commands.md): Every hop command in one list, with what each does: providers, chats, setup, projects, the workspace, skills, the dashboard and computers. - [Security](https://hoprig.com/docs/security.md): How the dashboard is protected: sign-in, passwords, tokens, git reads, the folders it never shows, and secrets. - [Is this allowed?](https://hoprig.com/docs/is-this-allowed.md): What Anthropic's documentation covers, what is yours to judge about several accounts and limits, and the sources. - [Where it runs](https://hoprig.com/docs/where-it-runs.md): Linux, macOS and Windows: what is run and tested, what is not, and what Hoprig needs. - [Not built, or not proven yet](https://hoprig.com/docs/not-proven-yet.md): Everything that is not built or not proven yet in this version, said plainly. ## For agents - [The Hoprig skill](https://hoprig.com/skill.md): A skill that teaches an AI coding tool to use Hoprig: check the setup, see providers and limits, continue a session elsewhere, look before acting. - [Tools: the commands and Tia's tools](https://hoprig.com/tools.md): Every Hoprig command, the ones that answer in JSON, and the tools of Tia, the assistant, with their confirmation levels. Made from the code at every build. ## Comparisons - [Hoprig and Claude Code's own apps and Remote Control](https://hoprig.com/compare/claude-code-remote-control.md): Claude Code's own apps reach a session from a phone and run work in the cloud. Hoprig sets up the computer, keeps a workspace in git and moves sessions. - [Hoprig and Happy](https://hoprig.com/compare/happy.md): Happy is a free, end-to-end encrypted phone app for Claude Code and Codex. Hoprig sets up the computer, keeps a workspace in git and moves sessions. - [Hoprig and the do-it-yourself way](https://hoprig.com/compare/tmux-ssh-vpn.md): tmux, SSH, a VPN and a terminal app put Claude Code on a phone with nothing new to trust. Hoprig adds machine setup, a workspace in git and provider moves. ## Optional - [Changelog](https://hoprig.com/changelog.md): What changed in each version of Hoprig, newest first, in plain words: new features, fixes, and what was not run yet on macOS. - [Compare](https://hoprig.com/compare.md): Hoprig next to Claude Code's own apps, Happy and the do-it-yourself way with tmux, SSH and a VPN: what each does, and what each does better. - [Docs](https://hoprig.com/docs.md): How to install Hoprig, code in a chat, carry on at a usage limit, keep a workspace in git, set up computers, link several of them, and what is not proven yet. - [Use cases](https://hoprig.com/use-cases.md): Who Hoprig is for, in thirteen cases, each with the problem, what Hoprig does step by step, what you need and what it does not do. - [Everything in one file](https://hoprig.com/llms-full.txt)