Remote control for AI coding agents. Rust + GPUI + QUIC/UDP. Available on iOS/Android
Zedra
An experimental remote code editor on mobile with GPU-accelerated rendering powered by Zed's GPUI, P2P tunnel over QUIC/UDP by Iroh. Zedra focuses on providing mobile-first experience for developers to read code, view changes, and run AI agents over a secure, P2P tunnel.

Security Note
Zedra is early, but it provides the most secure approach with e2e encryption, direct connection, and zero-trust model in a near future. If you have any concern about security, email me at tanle@zedra.dev, I will support to bring Zedra to your workflow with trust.
Download App
- iOS โ AppStore or TestFlight
- Android โ now on Google Play
Desktop daemon
Note: Zedra uses direct P2P connections when possible, but may fallback to relays if blocked by Symmetric NAT or CGNAT (common in home networks). Works best on LANs and supported relay regions. For high latency issues, please reach out. Learn more: How NAT traversal works
MacOS/Linux
curl -fsSL zedra.dev/install.sh | sh Install agent hooks for notification
zedra setup Start daemon in working directory
zedra start --detach
Windows
irm https://zedra.dev/install.ps1 | iex
zedra start --detach
Claude Code
# Config Zedra skills, hooks for Claude zedra setup claude In Claude Code, reload plugins and start Zedra
/zedra-start
Codex
# Config Zedra skills, hooks for Codex
zedra setup codex
In Codex, reload plugins and start Zedra
$zedra-start
Scan the QR code with the Zedra app. That's it.
Uninstall
zedra uninstall
Stop running daemons, remove agent hooks, delete local state (~/.config/zedra, ~/.cache/zedra), and delete the binary. Add --yes to accept all.
How It Works
zedra startruns a lightweight daemon on your desktop- Phone and desktop discover each other automatically โ direct P2P, relay fallback
- All traffic is encrypted end-to-end with TLS 1.3. No credentials leave your device
Status
Zedra is under active development. Core features are stable and in use โ bugs, rough edges, and breaking changes should be expected. Feedback and issues are welcome on GitHub.
License
MIT ยฉ Tan Le