Agent work does not arrive in neat one-minute increments. Some days you run a single session while you write. Other days Claude Code, Codex, Cursor, and Pi are all working at once and the callouts stack up. A plan that asks you to ration minutes punishes exactly the days the product is most useful.
Starting with version 2.4, Pro and Max both include unlimited callouts within a fair-use window. Pro is sized for regular agent work. Max gives heavy users five times the voice capacity in every window. And the two-way side of Agent Callout keeps expanding, so answering an agent from the notch covers more of the conversation.
Callouts stop being a meter
Both plans now work the way you do: use what the day needs, and let the plan absorb the bursts. Pro includes up to fifteen minutes of spoken audio in any rolling five-hour period, enough for steady sessions to narrate themselves all day. Max includes seventy-five minutes per window, exactly five times the capacity, sized for parallel agents and long Complete-style callouts.
If a window ever fills, voice callouts pause and resume on their own. Nothing to buy, nothing to manage, no decision to make mid-flow. The live session board, replies, and supported decisions stay fully available while the voice rests, and the Usage panel tells you exactly when it returns.
Answer more of the conversation from the notch
The reply side of Agent Callout keeps widening. Direct replies go back to the session that spoke. Claude planning questions arrive as focused cards with their choices laid out, so a multiple-choice question gets a one-click answer. Supported permission requests keep their clear Allow and Deny. All of it without hunting through terminal windows to find the session that asked.
Every answer is also accountable now. Agent Callout verifies the exact session pane before it delivers a reply, waits for the right moment when a session is mid-turn, and reports what actually happened: delivered, copied for you to paste, or expired. An answer you send from the notch is never a guess.
A Usage panel that tells the truth at a glance
Open Settings and the Usage panel now leads with what matters: Unlimited callouts, and a live picture of the current window. The bar shows how much of the window you have used, the caption says when it resets, and if callouts are paused it says exactly when they resume. No arithmetic, no countdown anxiety.
Your account page on the web shows the same truth with more room: the current window and the month side by side, each with its own reset time. The plan you chose and the usage you actually have, in one glance, matching what the app says because both read from the same source.
Product decision
The principle behind 2.4
Pricing should follow the shape of the work. Agent days are bursty, so a plan that meters every minute makes you ration attention at the exact moment your agents need it most. Fair-use windows keep the economics honest while letting you stop counting, and the higher tier is simply more capacity, not a different product.
The same standard applies to the notch. A reply you cannot trust to arrive is not a reply, and a usage number you have to decode is not information. Version 2.4 is the product telling the truth in both directions: your agents speak as much as the day demands, and every answer you send reports exactly what happened.