⌨️ Lancetta how-to #5 — let your agents check the room before a long task
⌨️ Lancetta’s advice isn’t only for the menu bar. Since 0.61.0 it’s also a command, and your agents can use it.
Three commands, installed with one click in Settings › General › Command line:
lancetta statusline
lancetta advice
lancetta usage --json
📟 statusline: one line for Claude Code’s status line or your prompt, both windows and then the suggestion that matters.
🧠 advice: the most urgent suggestion as plain text, or nothing. Perfect for a session-start hook, so every session begins knowing the room left.
🧾 usage --json: everything, for tmux, SketchyBar or a script.
🤖 And an Agent Skill for Claude Code and Codex: before a long task they check the room left and tell you in one sentence.
They read what Lancetta last saw, never an agent: no keychain, no dialog, no request to your account.
📖 The full walkthrough, hook and status line included: https://lancetta.substack.com/p/lancetta-how-to-5-let-your-agents
Free · universal binary · macOS 15+ · signed and notarized
👉 Try it / see what’s new: https://lancetta.app
💬 Questions and ideas: https://discord.gg/rdWu5yFCR6
What would you put in your own status line? 👀
Replies
This is actually a smart approach! Giving agents context before they start a task can save a lot of unnecessary back-and-forth. Love the idea of keeping it simple with CLI commands. 🚀
@ankit_pandit_dev Thanks 🙌 that was the idea, one command at session start instead of the agent finding out halfway through
I’d show when each usage window resets and when the reading was last refreshed. If it’s stale, the agent can flag that before starting a long task based on a limit that may have changed.
@al_kub yes, a stale reading is exactly what bites an agent before a long task... the JSON already carries both: resetsAt for each window and observedAt on every reading. The Agent Skill tells the agent that anything older than about 15 minutes means Lancetta isnt polling, so say so and don't treat the numbers as current. In that case advice prints nothing and statusline adds a "seen 20m ago"
the gap you spotted is the status line itself, it shows the percentages but not when each window resets. tbh that's half of what you need to decide whether to wait, so it goes on the list