Coding-agent guides and setup resources
Start with one agent and one repository. These public resources explain how to set up Bivy, check runtime support, and choose how to connect your machines. Reference documentation currently lives in the open-source repository on GitHub.
Run your first session
Follow the Bivy quickstart for installation and your first task. For your own-machine setup, have a Mac or Linux machine, a repository, and the agent’s model account or API key ready. Model-provider usage is separate from Bivy.
Before depending on a particular approval, resume, or fork behavior, check the agent support matrix. Claude Code, Codex, Pi, OpenCode, Grok and Kimi Code are the release-tested paths; capabilities vary by runtime.
Connect securely
- Security model: understand the boundaries between your execution machine, the relay, and your model provider before granting access to a repository or private service.
- Credential sync: learn which Bivy-managed keys and supported OAuth credentials can sync, and which native agent logins remain per-machine.
Remote access does not make a sleeping machine execute code. If work needs to progress while your laptop sleeps, run the agent on another machine that stays awake and online. Compare the ways to do that.
Choose hosting and workflow
- Run Claude Code or Codex from your phone: every current option, from the official Claude and ChatGPT apps to Happy, Paseo, Moshi, SSH and Bivy, compared by agent support and where the agent runs.
- Run Codex from your phone: Codex cloud and remote connections in the ChatGPT app, Happy, Paseo, Moshi, SSH and Bivy, compared by where Codex runs, which computers they support, and whether the computer must stay awake.
- Get a phone notification when Claude Code or Codex finishes: Remote Control push, the ChatGPT app, Claude Code hooks and Codex
notifywith ntfy, Happy, Moshi and Bivy, compared by which events they push and whether you can answer from the phone. - Keep Claude Code or Codex running after you close your laptop: keep the laptop awake, run the agent on a machine that stays on, or use Claude Code cloud sessions, Codex cloud or a Bivy cloud computer, compared by where the agent runs and what it can reach.
- Run Claude Code or Codex on a home server: a Linux box or a VM on Proxmox, TrueNAS or Unraid, headless sign-in for Claude Code and Codex, and tmux, Remote Control, Codex SSH hosts, Paseo and Bivy for reaching it from your phone.
- Claude Code hit its usage limit: what each limit message means, how to check your usage, and the options: wait for the reset, usage credits, an API key, another model, or another agent.
- Compare Bivy with hosted agents, IDE agents, and SSH: choose based on execution location, setup effort, interaction, and review—not just a feature count.
- Self-hosting quickstart: instructions for operating the app and relay yourself. Self-hosting the workspace is different from hosting model inference.
- Current Bivy Cloud plans: review the free allowance and paid option before starting.
Need help?
Browse the full public documentation or contact Bivy. When reporting setup trouble, include your OS, Bivy and agent versions, and a redacted error message. Never include API keys, OAuth tokens, or private repository contents.