Launching today
Lancetta
Know when to push your coding agents, and when to ease off
20 followers
Know when to push your coding agents, and when to ease off
20 followers
Lancetta is a native macOS menu-bar app that tells you when and how to use Claude Code and Codex. From every quota reading it says what to do next: spend a window about to reset unused, slow down before an agent stops, move to the other agent while it has room. It checks what your agents load at every start, frees the background processes they leave behind, and keeps every session on one wall, the ones waiting on you first. Reading quota spends zero tokens. Free, signed, macOS 15+.
Hi Product Hunt 👋
I run Claude Code and Codex all day, on several projects at once. Lancetta started as a quota number I could trust, and became the thing that tells me what to do with it.
😬 Why it exists
One afternoon my status line said 0% while the tool itself said 59% left: a three-hour-old reading, a refused turn that records no limits, and a reset time already in the past, stacked up into a number I'd have believed. The same day I found 28 orphaned agent processes in fourteen trees, holding 2.68 GB, twelve of the trees serving folders I'd already deleted.
🧭 A percentage tells you what's gone. Lancetta tells you what to do.
- ⏸️ Slow down — "At this pace Claude stops in 51m, then waits 1h18m for its reset": pause the sessions that can wait, lower the effort, give routine work to a lighter model, use a free reset if you hold one, or move the next tasks to Codex while it has room.
- 🚀 Push on — "Codex has room: 88% of this 5-hour window is unused, and it resets in 39m": quota you'd otherwise lose, so start the long task now.
- 📈 Where this pace lands — the projection drawn on the bar itself (or as charts, or as a clock): a ghost of where the window ends up at this rate, amber only when it runs out before the reset. Limits charts the week and the 5 hours, with a cone of where your rate leads.
- 🧹 Fix the setup — 22 checks of what Claude Code and Codex load before your first prompt (oversized instructions, skills past their budget, hooks and servers that can't start), with Fix All, or hand a finding to the agent itself.
- ♻️ Reclaim — the background trees the agents leave behind, which nothing else will ever close.
- 🗂️ Sessions — every Claude Code and Codex session on one wall, read from what the agents themselves report, with nothing installed into either: what waits on you, what's working, what you finished.
- ⌨️ Your agents ask too — a `lancetta` command and an Agent Skill for Claude Code and Codex: before a long task the agent checks how much room is left and what Lancetta suggests, in one sentence, and `lancetta statusline` puts the same line in your prompt. Installed only when you click it in Settings.
Every suggestion says what it rests on — the percentage, the rate over the last hour, when the window resets — and a banner comes only when it changes what you do next: an agent about to stop, room about to go unused, a free reset about to lapse.
🔍 And the numbers stay honest
- 💸 Reading quota costs zero tokens: twelve account reads moved the lifetime counter by exactly 0, while a day of work moved it by 29 million. No model is ever asked.
- ⏱️ Every reading carries its age; an unknown is a dash, never 0%.
- 🏷️ The bucket that refused you is named, never averaged away.
- 🧮 A total that's missing an agent says so ("Claude only", "Codex not read"), and a week compares against the week before it day by day, today left out.
- 🪟 The panel under the menu bar, the island under the notch, and a full window with usage over weeks, last week's recap and each agent in detail.
👥 Who it's for
Anyone who hits their 5-hour or weekly window and wants to know before it happens, and anyone juggling several agent sessions who keeps losing track of which one is waiting.
💶 Pricing
Free. Native macOS, universal binary, signed and notarized, in seven languages, macOS 15+.
🚫 What it will never do
Ask a model how much quota is left (it spends the very thing it measures), route around a limit, or keep a session warm on a timer.
I'd love your feedback: which suggestion would actually change how you work — and what should Lancetta tell you that it doesn't yet? 🙏
— Giovambattista
https://lancetta.app · 💬 https://discord.gg/rdWu5yFCR6
Showing a dash when the quota is unknown is a good call. A stale 0% would send someone off to troubleshoot the wrong thing.
@oleksii_sekundant thanks, that was the whole point. A dash tells you less than a number, but a wrong number is worse imo