One Model for Every CLI Agent

Terminal-based agents all share the same shape: a long-running process that prints output and waits for your input. CLIAnywhere turns that process into something you can reach from any device.

The agent is a terminal process

Claude Code, Codex, Gemini CLI, OpenCode, and Aider all run inside a shell on your computer. They read your prompts from stdin and print edits, diffs, and test results to the terminal.

Agent tasks are long

Multi-file refactors, dependency upgrades, test-suite runs, and training jobs routinely take minutes to hours. Nobody wants to sit at the terminal for the whole run.

Agents need check-ins

Agents stop to ask questions, hit failures, or finish a phase while you are away. Being able to look in and reply — without running back to the desk — is the whole game.

CLIAnywhere is the window

The terminal session stays on your computer; the app on your phone or browser is a live window onto it. Watch, scroll back, and type — from wherever you are.

agent — 80x24
~/webapp $ codex "add rate limiting to the API"
● Reading openapi.yaml and middleware/
● Proposing a plan (4 steps)
✻ Working…
● Applied edits to middleware/ratelimit.go
● Ran tests — 1 package failed
● Fixed the failure and re-ran — all tests pass
> looks good — commit it
● Committed on branch ratelimit
~/webapp

Illustrative session: a terminal-based coding agent, viewed in the CLIAnywhere mobile app.

Any Agent That Lives in a Terminal

CLIAnywhere has no per-agent integrations, plugins, or vendor partnerships — and needs none. If your agent runs in a terminal on your computer, remote access to that terminal session is all it takes.

claude

Claude Code — Anthropic's terminal-based coding agent, also covered in its own use-case page.

codex

OpenAI's command-line coding agent. Run it in your shell and reach the session from your phone.

gemini

Google's Gemini CLI agent for terminal-based coding workflows.

opencode

Open-source terminal coding agent. Same rule: it runs in a shell, so it is reachable.

aider

Open-source AI pair programming in the terminal. Long refactors are its normal workload.

your-agent

A custom script, a model training run, a build pipeline — anything printing to a terminal works the same way.

Claude Code, Codex, Gemini CLI, OpenCode, and Aider are products of their respective owners. CLIAnywhere is an independent product that is not affiliated with, endorsed by, or partnered with Anthropic, OpenAI, or Google. It does not run, host, or control any agent — it provides remote access to the terminal session on your computer.

Why Persistent Sessions Matter for Long Agent Tasks

Builds, training runs, and multi-hour agent tasks fail in one specific way: the connection between you and the terminal drops. Persistent sessions remove that failure mode.

Agent runs outlive your attention

A long refactor, an upgrade across every dependency, or a training job can run for hours. With CLIAnywhere you check in when it matters instead of babysitting the terminal.

Disconnects change nothing

The session runs on your computer, not on your phone. Lock the phone, close the app, switch from Wi-Fi to cellular — reconnect later and the scrollback, the shell state, and the agent are all still there.

No SSH-plus-tmux rig required

The classic fix is tmux or screen behind an SSH tunnel. CLIAnywhere gives you the same continuity across desktop and mobile out of the box, with no server configuration or port forwarding.

Honest limits: CLIAnywhere does not start agents for you, does not host them, and cannot keep an agent alive if your computer shuts down, reboots, or goes to sleep — keep the machine running. It provides remote access to a terminal session, nothing more and nothing less.

An Agent Run That Outlives Your Evening

One pass through the workflow: start a long task before you leave, steer it from your phone, collect the results later.

It is 6 p.m. and you want an agent to migrate a service from one API client to another before tomorrow — a task that touches dozens of files and needs the full test suite green at the end. You start it in a terminal on your office machine, confirm it has understood the task, and leave for the day.

  1. From the train, you open CLIAnywhere and connect to that terminal session. The agent is halfway through the migration; you scroll back through the files it has edited so far.
  2. It stops with a question about a deprecated endpoint. You read the context on your phone, type the answer, and the run continues.
  3. A test run fails an hour later. You are already home; you watch the agent retry with a fix, and the suite goes green.
  4. You collect the results. The built-in file browser pulls the migration report and the changed diff down to your phone over the same encrypted channel.

Nothing about the agent was hosted or proxied by CLIAnywhere — the run lived in a terminal on your computer the whole time. You just never had to be in front of it.

Agent Output, Encrypted End-to-End

Everything that crosses the wire — keystrokes, agent output, downloaded files — is encrypted end-to-end: keys are negotiated between your devices with SPAKE2, and payloads are sealed with AES-256-GCM, so the relay used as a fallback only ever sees ciphertext. Every computer you link is protected by a Security Code generated locally on that machine, never uploaded. See the security overview for details.

What CLIAnywhere Supports

Platform and capability facts for this use case, all verified against the shipping product.

Windows hostYes
macOS hostYes (signed and notarized, distributed via GitHub Releases)
Linux hostYes (desktop UI or headless CLI build)
iOS clientAvailable on the App Store — see the download page
Android clientGoogle Play
Web clientwebapp.clianywhere.com
Shellsbash, zsh, fish, sh, cmd, pwsh, powershell
Persistent sessionsYes
Desktop↔mobile session switchingYes
File browser + downloadYes
P2P direct (DTLS)Yes
Encrypted relay fallback (WSS)Yes
End-to-end encryptionSPAKE2 + AES-256-GCM
Port forwarding / public IPNot required
Open-source daemongithub.com/CLIAnywhere/clianywhere_daemon

Frequently Asked Questions

Yes. Codex CLI runs in a terminal on your computer, and CLIAnywhere gives you remote access to that terminal session from an iPhone, Android device, or web browser. Your phone shows the same live output and accepts the same input as the terminal on your desk. There is no special Codex integration and CLIAnywhere is not affiliated with OpenAI — any program running in your terminal, including Codex, becomes reachable this way.
Leave Codex running in a terminal on your computer and open that session in the CLIAnywhere app. You can watch its output live, scroll back through everything it did while you were away, and reply when it asks a question or hits an error. Because the session is persistent, monitoring survives phone lockouts and dropped mobile connections — you just reconnect and the output is still there.
If the agent runs in a terminal on your computer, point CLIAnywhere at that terminal session. From your phone you can watch output as it streams, read the diffs and test results the agent prints, and type replies when it pauses for input. This works for any terminal-based agent — Claude Code, Codex, Gemini CLI, OpenCode, Aider — because CLIAnywhere provides remote access to the terminal, not a per-agent integration.
Keep the computer on and let the agent run inside a CLIAnywhere terminal session. The session is persistent: it survives the app closing, your phone locking, and mobile connections dropping, because the process runs on your computer. What CLIAnywhere cannot do is keep an agent alive through a shutdown, reboot, or sleep — it provides remote terminal access, not a hosted agent runtime.
Connect to the computer the agent is running on with the CLIAnywhere app and open its terminal session. Long-running tasks are the normal case: the session collects hours of scrollback while you are away, and you can reconnect from any device — phone, tablet, or browser — to check progress, review what happened, or send the next instruction. No port forwarding or public IP address is needed.

Keep Reading

Give your agents a longer leash

Free download, one-minute setup, and your terminal session on every device.

Download App Setup Guide