Show HN: Omnara – Run Claude Code from anywhere

Original Article ↗ Hacker News Discussion ↗

DIY vs SaaS and “Why Pay?”

  • Many argue Omnara has no strong moat: you can already do this with SSH + tmux/screen/byobu, Tailscale, Termux, VNC, or Vibeltunnel-style tools, often fully self-hosted and free.
  • Counterpoint: this is true of most software; people still pay for convenience, support, polished UX, and not having to build/maintain their own stacks.
  • Some describe quickly “vibe-coding” bespoke tools (e.g. helpdesks, support bots) and preferring custom fits over SaaS, while others warn that maintenance, infra, and bug-fixing remain real burdens even with LLMs.

Target Users and Workflow Fit

  • Founders position the product mainly for users who want zero setup and seamless cross-device continuity (terminal ↔ web ↔ mobile), including non-technical “vibe coders.”
  • Several developers say mobile Claude Code is genuinely useful: kick off long tasks, get push notifications, review/respond while commuting or away from the desk.
  • Skeptics question feasibility: proper QA, code review, and running apps are hard on phones; their bottleneck isn’t “waiting for agents” but validating outputs.

Privacy, Security, and Self‑Hosting

  • Omnara routes all messages through its servers to support sync and notifications; chats are stored server-side and deleted immediately on user deletion.
  • Backend is open source; mobile/web open-sourcing and possibly self-hosting are on the roadmap, with some users explicitly needing on‑prem / custom API endpoints for enterprise approval.
  • Multiple commenters are uncomfortable sending proprietary code through yet another third-party, preferring local-only or peer‑to‑peer (Tailscale/ngrok/VNC) setups.
  • Questions raised about what data is collected, regulatory compliance, and risks of central servers being a high‑value target.

Future of Agentic Development

  • Strong enthusiasm for the broader pattern: humans set goals and manage agents; coding shifts toward task orchestration, with agents doing most implementation.
  • Some envision managing teams of specialized subagents and already report running long unsupervised Claude Code sessions using structured workflows (PLAN.md, subagent orchestration, strict checks).
  • Others report brittle, duplicate, or incoherent code when agents run too long unsupervised; techniques like “red team vs blue team” agents and subagents are suggested to mitigate this.
  • Debate over job impact: some see this empowering problem-solvers; others worry C‑suites will use it to reduce headcount or commoditize “codemonkey” work.

Competition, Platform Gaps, and UX Issues

  • Concern that Anthropic or IDE vendors will ship similar cross-device agent experiences, eroding Omnara’s value; response is to differentiate via multi-agent, multi-IDE support and richer components.
  • Android support exists but is delayed in the Play Store; Windows support is blocked by terminal library issues. Users also report authentication hurdles, copy/paste limitations on iOS, and landing-page glitches.
  • Some criticize complex TUI-style CLIs (including Claude Code) as hard to wrap; they wish more tools exposed simple JSON/text protocols for multi-UI frontends.