Hapi
Hapi is a local-first remote-control shell for official coding CLIs. Rechecked 2026-08-17: github.com/tiann/hapi had 4,778 stars, license AGPL-3.0. README says it runs Claude Code, Codex, Cursor Agent, Grok Build, OpenCode, Kimi, Copilot, Antigravity, and Pi locally, then hands the session to a Web / PWA / Telegram Mini App. It is not its own frontier model and not a Happy rename. The repo calls itself a local-first alternative to Happy. We found no public dollar table on hapi.run.
Compare Happy if you wanted the slopus/happy client, Claude Code or Codex CLI if you wanted the wrapped agents themselves, and OpenCode or Kimi Code if you already live in those CLIs.
Key Features
- Wraps official agents: Native-first copy says it does not replace the terminal agent. Same session, optional remote handoff, no restart.
- Install is npx, not a vibe SaaS: README starts with
npx @twsxtd/hapi hub --relayandnpx @twsxtd/hapi. The terminal prints a URL and QR code.hapi serveris an alias. - Relay is named: WireGuard + TLS end-to-end encryption on the bundled relay. Self-host options point at Cloudflare Tunnel or Tailscale in the install doc.
- AFK approvals and voice: Phone tap to approve, built-in voice assistant, optional workspace browser via
hapi runner start --workspace-root.
Use Cases
- People who already pay for Claude Code or Codex and want phone control without leaving the official CLI.
- People comparing Hapi vs Happy after they see the "Why not Happy" doc.
- People who will self-host instead of a hosted coding agent.
Limitation: Homepage HTML is a client-rendered shell with almost no static pricing text. We did not invent a Pro seat. Agent quality is the wrapped CLI's quality.
Pricing
github.com/tiann/hapi and hapi.run on 2026-08-17.
| Piece | Price | Notes from first-party pages |
|---|---|---|
| Hapi app / CLI | No public dollar row | AGPL-3.0. 4,778 stars. |
| Wrapped agents | Their own seats | Claude Code, Codex, Cursor Agent, and the other listed CLIs. |
| Relay / self-host | Recheck install docs | Bundled E2E relay, or Cloudflare Tunnel / Tailscale. |
If a leftover blog still sells Hapi as a priced vibe-coding SaaS, treat this table as the source.
Getting Started
- Run
npx @twsxtd/hapi hub --relay, thennpx @twsxtd/hapi. - Scan the QR code or open the printed URL.
- Recheck Happy if you wanted that client instead.
- Do not invent a Hapi monthly seat.
Frequently Asked Questions
Is Hapi a new coding model?
No. It wraps official CLIs you already have.
Same as Happy?
No. The README says it is a local-first alternative and points at docs/guide/why-hapi.md.
Is there a $20 Pro plan?
Not on the public pages we opened.
Alternatives
- Happy: slopus/happy mobile/web client for Claude Code.
- Claude Code: The Anthropic CLI Hapi can wrap.
- Codex CLI: OpenAI's terminal agent, also on the wrap list.
Tips
- Quote AGPL-3.0 and 4,778 stars from the GitHub check.
- Quote the wrap list, not a new model name.
- Recheck Happy before you treat Hapi as the only remote client.
Conclusion
Hapi is an AGPL local shell around official coding CLIs, not a priced vibe-coding model. Start at github.com/tiann/hapi, then decide whether Happy or the wrapped CLI already covers the remote loop you need.
Comments
No comments yet. Be the first to comment!
Related Tools
Codebuff
www.codebuff.com
YC terminal agent. Rechecked: npm i -g codebuff. Public seats $100 / $200 / $500 plus 1¢/credit. Not a $20 Cursor twin.
Tabnine
www.tabnine.com
AI code completion and chat. Rechecked: public tabnine.com/pricing returned 403. Do not invent Dev $14 or Pro $20. Not a 2024 Atom/Sublime blurb.
Amp
ampcode.com
Sourcegraph agent. Rechecked: Megawatt $20 / Gigawatt $200. Not a Cody rename; not $10/day unconstrained Opus.
Related Insights
Skills + Hooks + Plugins: How Anthropic Redefined AI Coding Tool Extensibility
An in-depth analysis of Claude Code's trinity architecture of Skills, Hooks, and Plugins. Explore why this design is more advanced than GitHub Copilot and Cursor, and how it redefines AI coding tool extensibility through open standards.
Seven AI Coding CLIs, Six Months: No Matter How Strong the Model, Work Needs Supervision
Claude Code, Codex, opencode, pi, omp and DeepSeek Harness all have personalities. After six months of deep use I run a division of labor: pi for the fastest cheapest reviews, omp for complex PRs, DeepSeek Harness on V4 Flash for high-frequency low-cost review, and Claude Code, Qoder and Cursor for writing. No matter how strong the model, work needs supervision — ideally from an independent third party.
Claudesidian: Transform Obsidian into an AI-Powered Second Brain
Discover Claudesidian, an open-source project that perfectly integrates Obsidian with Claude Code. Built-in PARA method, custom commands, and automated workflows for a complete idea-to-implementation solution.