Tailscode vs Pocket AI Lab

Send a prompt to Claude Code, opencode or Oh My Pi from your phone, pocket it, and watch the turn finish on your Lock Screen. Approve the risky command, answer the agent's question, read every diff — from anywhere on your own Tailscale network. WHAT YOU NEED A Mac or Linux machine on your tailnet (Tailscale is free for personal use) running one of your agents: opencode, Claude Code through the open-source claude-bridge, or Oh My Pi through the open-source omp-bridge. Set-up is three verified steps, and a built-in demo lets you try every screen before you wire anything. THE TURN, LIVE ON YOUR LOCK SCREEN • A Live Activity per conversation: what it is doing, for how long, and how many tools it has run. • An approval request alerts you; a finished turn lingers with its outcome. Tap to land in the exact chat. • Updates arrive by push while the app is suspended. A NATIVE TRANSCRIPT, NOT A TERMINAL • Streaming text written at the model's own speed, with thinking and tool runs folded into readable groups. • Diffs, commands and their real output inline; syntax-highlighted code with byte-exact copy. • Allow once, always, or deny — without leaving the chat. Agent questions become tappable cards. • Queue the next message while a turn runs, edit it before it goes, or stop the turn on the server. • Every picture the agent hands you, in a gallery you can zoom, save or share. EVERY MACHINE ON ONE BOARD • All your servers' conversations in one list, live status first, grouped by project. • Read the repository before you trust the change: branch, staged, changed, untracked, and the diff of any file. • Subagents dock inside the turn that spawned them; a compaction is drawn as a seam with what it kept. • Slash commands from the server's own catalog, per-message model and reasoning effort, forking. KNOW WHAT IT COSTS • Every conversation priced, every month totalled across machines — by model, project and tool. • Plan limits and quotas for Claude, Grok, Ollama Cloud, DeepSeek and more, on Home and in a widget. YOURS, END TO END • Point-to-point over your tailnet. No relay servers, no account, no analytics, nobody reading your code. • Programmatic UIKit, Swift 6, Liquid Glass on iOS 26. Dark mode, Dynamic Type, haptics. • Open source under GPL-3.0 — the app, the engine, and both bridges. FREE, WITH AN HONEST PRO The free app is the complete core: one server, full chat, approvals, attachments, models, forking, and Live Activities. Tailscode Pro is a one-time purchase, no subscription: multiple saved servers, concurrent Live Activities, and a supporter badge. Tailscode is independent open-source software: not affiliated with or endorsed by Anthropic, Tailscale Inc., the opencode project or Oh My Pi. Privacy policy: https://midgarcorp.cc/tailscode/privacy Source code: https://github.com/guitaripod/Tailscode

Pocket AI Lab scans your device, tells you honestly which models will actually fit, downloads them straight from Hugging Face, and lets you chat with them fully offline. Multi-backend inference — MLX, llama.cpp (GGUF), Core ML, and Apple Intelligence, switchable at runtime Multimodal — text, images and video frames (vision models), speech-to-text (Whisper) Import any model by link — paste a Hugging Face URL, pick a quantization, get a fit verdict before a single byte downloads Honest memory budgets — recommendations are computed from the real per-process memory limits of your device, not marketing RAM numbers Live metrics — CPU, RAM and thermal pressure while the model is thinking Streaming chat — token-by-token, with full conversation history on every backend