Ship Claude Code sessions in parallel.
Agent Tile puts any number of agents on one macOS canvas:
• ⌘-click → broadcast typing to a set
• ⌘⌘ → fisheye any panel
• Crash? Relaunch restores every session
Full source. $10 once.
https://t.co/lJaxUhahP4
Running 5+ AI coding agents at once and losing track of which window is which? Agent Tile tiles them on one screen with per-project color, name and live status. One glance, not five.
⌘-click a set of panels, then type once — it broadcasts to all of them. Kick off the same command across every repo at the same time.
Agent Tile for macOS. Full source, $10 once.
https://t.co/lJaxUhahP4
Nobody warns you that running many agents makes you a manager. Same skills: clear instructions, checking in, unblocking — and not micromanaging the ones that are fine.
Six terminal tabs, all named 'zsh', each running a different agent. You alt-tab, guess, land in the wrong one. Agent Tile gives each its own tile, color and name. I built it because I kept doing exactly that.
Parallel agents change how you scope work. You stop asking 'what's my next task' and start asking 'what are my next five, and which can run unattended.'
Hot take: the bottleneck in AI-assisted coding stopped being the model months ago. It's you, context-switching between six sessions, being the slow part.
No subscription, no login, no telemetry. Agent Tile is $10 once with full source — fork it, rebrand it, wire in your own tools. https://t.co/lJaxUhahP4
One prompt, every repo. ⌘-click the tiles you want, type once, and it broadcasts to all of them — run the same migration or test suite across your whole fleet in one shot. Agent Tile, macOS.
The feature buyers ask about most: crash-proof restore. Kill Agent Tile mid-work, relaunch — every panel returns in its own folder, resuming its own Claude session. It’s `claude --resume` for a fleet.
Reviewing AI-written diffs is a different muscle than writing code. Half the job is reading fast and saying “no, back up.” At 5+ agents that review time is the real constraint — not the model.
Shipped this week: ⌘↑/↓ scrollback that works INSIDE Claude Code's TUI (protocol-correct mouse reports), UTF-8 fixes for CJK, sync-typing to selected tiles.
$10 gets you every update + full source. https://t.co/lJaxUhahP4
Terminal managers treat every shell the same. But an AI agent session isn't a shell — it has states: thinking, waiting for you, done, stuck.
Agent Tile shows those states per tile so you only look when needed. Attention is the bottleneck, not compute.
Agent Tile started as my own pain: I run a one-person company where Claude Code sessions are the employees.
The board IS the office. Each tile has an avatar, a status light, and a job.
Build-in-public continues — what should the tiles show next?
The feature I use most: double-tap Cmd and the active terminal becomes a full-screen log you can scroll, search, and type into. Double-tap again — you're back on the board.
Small thing. Saves me 100 window switches a day.
Quit the app. Reopen. Every Claude Code session is exactly where you left it — same folders, same output, auto-resumed.
Agent Tile keeps the PTYs alive outside the UI, so a reload never kills your agents.
$10 buy-once: https://t.co/lJaxUhahP4