When you would want it

Long-running terminal tasks should not need you to watch the clock and return only to type one word.

Where to find it

All Features → Continue terminal AI after limits
Turn continuation on or off
Turn on or off
Settings

What you get out of it

A long task that hits a limit at midnight is still moving at half past, without you sitting up for it.
It works with whichever CLI you run - Claude Code, Codex, Grok, Kimi and the others are all recognised.
The retry cap means a repeatedly limited agent stops rather than looping all night.
Only the terminal output already on your screen is read, and only for a limit message with a stated retry time. An unclear message stops safely.

How to use it, step by step

Right-click a terminal running an agent and choose Continue after usage limits, or turn it on in AI settings.
Set how many continuations you will allow in one session - one, three, or five.
Work as normal. When the agent says it has hit a usage or rate limit and names a retry time, Agentfly waits for it.
At that time Agentfly types a single word - continue - and tells you it did.
Touch the terminal at any point and the pending continuation is cancelled, because you have taken over.

Why this beats the usual way

Everywhere else this is a shell script you wrote yourself, guessing at the retry time and re-sending blindly. Agentfly reads the agent's own stated reset time, sends exactly one word, caps the retries, and cancels the moment you touch the keyboard.

Find it in Universal Search

Continue terminal AI after limits · The assistant · usage limit · rate limit · resume after limit · keep going · unattended · grok · kimi · codex · claude code