Omnara – Remote Claude Code and Codex From Any Device

TL;DR

TL;DR: Omnara exposes Claude Code and Codex sessions through a web and mobile interface, letting you interact with coding agents remotely without sacrificing the power of local compute.

Source and Accuracy Notes

• Official site: omnara.com
• YC Launch: HN Thread (147 points)
• Product: web + mobile interface for Claude Code and Codex

What Is Omnara?

When you run Claude Code or OpenAI Codex locally, you’re tied to your machine. Omnara bridges that gap — it runs the agent on your own hardware but gives you a web and mobile interface to monitor, interact with, and control the session.

The agents still execute on your local machine (keeping your API keys and code secure), but you can check progress, provide guidance, and manage tasks from a phone or a browser tab on another computer. It is a remote control layer for local coding agents.

Setup Workflow

Prerequisites

• Claude Code installed locally (npm install -g @anthropic-ai/claude-code) or Codex configured
• A device to serve as the agent host (laptop, desktop, server)
• Web browser or mobile app for the Omnara interface

Step 1: Create an Omnara Account

Sign up at omnara.com. The service is free during beta.

Step 2: Install the Omnara Agent

On your host machine, install the Omnara agent and link it to your account:

npm install -g @omnara/agent
omnara login

Step 3: Start an Agent Session

# Start a Claude Code session
omnara start claude --project ~/my-project

# Or start a Codex session
omnara start codex --project ~/my-project

The agent runs locally. Omnara creates a secure tunnel so your web and mobile sessions can communicate with it.

Step 4: Access From Web or Mobile

Open app.omnara.com on any browser, or download the Omnara mobile app. Your running sessions appear as active cards — tap one to send messages, review file changes, or pause the agent.

Deeper Analysis

Why Run Agents Remotely?

The compute for running a coding agent is cheap — an hour of Claude Code usage might cost a few cents in API credits. But your time is expensive. Omnara solves the “I want an agent running on a big codebase while I’m on the train” problem.

Instead of SSHing into a remote server (and losing the interactive IDE context), you get a purpose-built UI for agent interaction. The agent still uses your local environment, which means it has access to your full repo, your installed tools, and your personal API quotas.

Security Model

The agent runs on your machine. Omnara does not store your code or API keys — it only brokers the session tunnel. Your Claude Code or Codex session authenticates directly with Anthropic or OpenAI using credentials stored in your local environment (~/.claude.json, environment variables).

Architecture

Your Machine (host)
  └─ Omnara Agent (local process)
       ├─ Claude Code / Codex (local)
       └─ Omnara Tunnel (secure websocket)
            └─ Omnara Cloud
                 └─ Web / Mobile Client (anywhere)

Practical Evaluation Checklist

• Agent runs on local hardware with full repo access
• Web UI for monitoring and interacting with sessions
• Mobile app available for on-the-go oversight
• No code or credentials stored on Omnara servers
• Secure tunnel communication between client and host
• Free during beta

Security Notes

• API credentials stay on your machine
• Session tunnel is encrypted
• Omnara acts as a relay, not a store
• Review agent file changes before accepting them from mobile

FAQ

Q: Does Omnara work behind a NAT or firewall?

A: The Omnara agent creates an outbound tunnel to Omnara’s servers, so it works even if your machine is on a restrictive network. You do not need to open inbound ports.

Q: Can I use my own API keys?

A: Yes. Claude Code and Codex use whatever credentials are configured in your local environment. Omnara does not inject or replace them.

Q: How is this different from VS Code’s remote SSH extension?

A: VS Code Remote SSH gives you a full IDE on a remote machine. Omnara gives you a purpose-built agent interaction UI for coding agents that run locally. The agents are not running in Omnara’s infrastructure — they are running on your hardware.

Q: Does it work with both Claude Code and Codex?

A: Yes. Omnara supports both. You can run one or the other, or multiple sessions simultaneously.

Conclusion

Omnara solves a real workflow gap: local coding agents are powerful but tethered to your desk. By decoupling the agent execution from your physical location, it enables a new interaction pattern — start a long-running agent session at your office workstation, then monitor and guide it from your phone during your commute. The agent stays on your hardware, your credentials stay local, and you get a clean mobile-first interface for staying in the loop. Worth trying if you regularly run coding agents across multiple contexts.