Lochan vs Model Pulse

Your coding agents keep working when you step away. Lochan is how you keep up with them, calmly. When Claude Code, Codex or any other agent on your computer stops to ask a question, open its real terminal on your phone, read what it did, and answer. Then get back to your walk. A CALM VIEW OF A BUSY HERD Every agent on the machine at a glance, with a line on where each one left off. The ones waiting on you always come first. - Tiles: the whole herd as a grid that moves as your agents do - Board: Blocked, Working, Done and Idle as columns you swipe through - Projects: agents grouped by the project they are working in ASK THE SHEPHERD One question, the whole herd. The Shepherd is an agent on your computer that reads every other pane. Ask it what needs you, what changed or whether anything is stuck, and it answers from what is on the screens, with a link to each pane it mentions. It runs Claude Code by default, or Codex, Pi, OpenCode or Grok. ANY AGENT CAN JOIN THE HERD Lochan does not plug into your agent. It opens the terminal the agent is already running in, so it works with Claude Code, Codex, Gemini CLI, Grok, OpenCode, Aider and anything else with a prompt. A new agent works the day it ships. A REAL TERMINAL, IN YOUR POCKET Tap any agent for the actual session: full output, scrollback and a keyboard. MAKE THE KEYS YOUR OWN Two rows of keys sit above the terminal. Drag any key to either row, switch off the ones you never use, and add your own: - Text keys for the commands you repeat, like /compact, /usage or /model sonnet - Chord keys for Ctrl, Option and Shift combos, like your tmux prefix or Ctrl-R SHARE A SCREENSHOT WITH YOUR AGENT Pick a photo, take one of the screen in front of you, or send a file. It lands on your computer and its path goes into the prompt, ready for you to finish and send. SAY IT. CHECK IT. SEND IT. Dictate a long instruction, read the transcript, and send it only when it reads right. Or write it in a real text box, with a draft that waits if you get pulled away. RUN THE HERD FROM YOUR PHONE Start a new agent in any project, rename it, and close it when it is done. BUILT FOR A PHONE Lochan tells each agent it is being driven from a phone, so it leads with the answer, keeps replies short and batches its questions. PAIRED IN ONE SCAN Run one line on your Mac or Linux machine, then scan the code it shows. Your phone makes its own key, and it stays in the phone's Keychain. Nothing to type, no private key to copy. Save every machine you own and switch between them from the title bar. No account, and no Lochan server in between: your phone talks to your computers over SSH. Away from home you need a route to them, and Tailscale is the two-minute version. ALSO IN LOCHAN: A PRIVATE AI CHAT Chat with Apple's on-device model, open models downloaded to your phone, or cloud models with your own API key. Conversations branch instead of overwriting, and you can share one with a link we cannot read. Lochan works with Herdr, the terminal workspace for AI coding agents, on the computers you own.

Model Pulse puts your AI subscriptions and API spend on your iPhone Home Screen, so you always know how much runway you have left. HOME SCREEN WIDGETS - Small, medium, and large widgets - See 5-hour and weekly plan windows, monthly spend, and token burn at a glance - Bar, dot, or dash limit styles to match your setup SUBSCRIPTIONS - Track plan usage for ChatGPT, Claude, Gemini, and more - 5-hour and weekly rate-limit windows with reset times - Sign in securely — tokens are stored in the iOS Keychain and never leave your device except to talk to the provider API SPEND - Month-to-date spend for OpenAI and Anthropic organization accounts - 24-hour token and request tracking - Optional monthly budgets to pace prepaid credits USAGE ALERTS - Optional notifications when a rate-limit window or API budget crosses thresholds you choose - Alerts keep working in the background — no need to keep the app open YOUR DATA STAYS YOURS Model Pulse has no accounts, no analytics, and no tracking. Keys live in the iOS Keychain. Requests go straight from your device to the providers you connect — there is no Model Pulse server.