Preparing your next view.
Release notes
Every feature, fix, change, and removal, posted here as it ships. This is a running log, not a highlight reel — the small stuff is on here too.
Calls now use the faster-responding voice engine unless you've chosen otherwise, so there's a much shorter pause before each reply. If you prefer the richer, more expressive delivery, choose Most natural under Voice engine in Personalization.
Personalization now lets you choose a faster-responding voice engine that cuts the pause before each reply, set how expressive the voice sounds, decide how patiently your caller waits for the other person to finish, and pick how much it says per turn.
A new preview button in Personalization plays your caller's opening line with the voice, pace, tone, expressiveness and pronunciations you've chosen, including changes you haven't saved yet, so you can hear exactly how it will sound.
Tell your caller what it should always know, like a membership under your name, a preferred time of day, or who to ask for. Your caller uses whatever is relevant to each call, and these notes never override your sharing and spending permissions.
When someone asks your caller to hold, it now says it will wait and stays quiet instead of asking whether they're still there. It also responds more naturally when a business suggests a different time, and spells names and codes more clearly over the phone.
Transcripts now recognise the business you're calling, your name, and any names you've added under Pronunciation more reliably, so the details you read back after a call are the ones that were actually said.
Avery, Maya and Claire are now available in Personalization, with an audio preview to help you choose. Your conversation tone, speaking pace and pronunciation preferences continue to shape whichever voice you select.
The choices you save under what a call may share — your name, your callback number, your address — and your fee-without-approval limit were not being carried into new calls, which used the standard defaults instead. Every call created from now on follows what you have saved, on web, mobile and through a connected assistant. Calls and routines created before today keep the permissions they were originally given; re-create a routine to pick up your current settings.
A time like 4:00pm was sometimes read out literally, as "four oh oh p.m." Times, prices, dates, phone numbers and confirmation codes are now spoken the way a person would say them out loud, on every English call.
Longer replies used to land as a series of separate sentences, each starting from a flat tone. The caller now carries its intonation through to the end of a thought, which makes a multi-sentence answer sound noticeably more natural.
You can now add names your caller gets wrong — a surname, a business, a street — along with the spelling that makes it come out right. Corrections apply to every call, in every language, and you can save up to 25 of them in Preferences.
A new speaking pace setting lets you pick a slower, clearer delivery for calls where the other person is writing details down, or a brisker one for short routine calls. It sits alongside conversation tone in Preferences.
Account limits now apply consistently whether you work in the web app or through a connected assistant, so heavy use on one never affects the other. If you do reach a limit, the message now tells you which action is affected and roughly how long to wait.
Account and calling protections now hold up more reliably against automated traffic, and each account's limits are its own rather than shared with everyone on the same network.
A live call that received something it could not interpret could previously affect calls running at the same time. Each call is now fully isolated, and an unreadable moment costs that one call a beat instead.
Searching your support requests for a phrase containing punctuation or symbols now returns the right results instead of failing or matching far too much.
Updating one preference through a connected MCP assistant no longer clears the other settings saved alongside it, and an unsupported call language is now rejected the same way it is in the app.
Times, prices and dates read out on a call now sound the way you would say them out loud — "six p.m." rather than a literal reading of the digits on the page. This applies to every call, in every language ExeclyAI speaks.
Calls now carry more of the natural rise and fall of a real speaker instead of a flat, recited delivery. The conversation tone you pick — warm, professional or direct — shapes it further, so the difference between them is easier to hear.
Switching to an authenticator app to retrieve a code no longer interrupts the sign-in session before the code is submitted.
Set up a routine once — the weekly supplier check, the monthly refill, the Friday table — and ExeclyAI places the call on your schedule from then on. Every run still gets its own plan, its own permissions, and its own report, and you can pause a routine at any time.
A busy signal or an unanswered ring is not an answer. You can now let a call retry up to five times, spaced as far apart as you like, and change your mind while the first attempt is still ringing. A call that reached a person is never redialled.
Finished calls now have a quick rating with optional tags and a note, so you can say what went right or wrong while it is fresh. We read them.
See what ExeclyAI has handled for you: the phone time you skipped, how many requests came back resolved, which businesses you call most, when you tend to need calls made, and how you have been rating them.
The prompts that ask you to confirm something now open centred on screen with a clear heading, plainer wording about what will happen, and buttons that name the action instead of a generic OK.
Replaced native browser confirmation popups with custom, in-app dialogs that match the rest of the application's design.
Fixed an issue that prevented some accounts from clearing their finished calls from the history tab.
Calls now use a dedicated voice engine for speech, with a voice matched to the call's language and an optional voice picker in your calling preferences. It sounds noticeably more natural than before.
Adjusted subscription pricing across plans. Current pricing is always shown before you subscribe or change plans.
Outbound calls now go out from whichever of our business phone numbers is closest to the destination, instead of always using the same one.
Added a live status page showing uptime for every part of the service, plus a way to report a bug directly from the app.
Some requests are now held for a quick human review before a call is placed, as an added safeguard for edge cases automated screening isn't confident about.
ExeclyAI can carry on a call in English, Spanish, French, German, Italian, Portuguese, Dutch, Polish, Russian, Arabic, Hindi, Chinese, Japanese, or Korean, matched to what the task needs.
Support requests now go through proper case tracking with a searchable Help Center, instead of email-only support.
The free trial plan is no longer offered to new signups. Existing accounts already on it are unaffected until they choose to upgrade.
Power plan accounts can connect ExeclyAI as a tool inside Claude and other MCP-compatible assistants, so tasks can be created from outside the app.
Added a library for saving call instructions and reusing them on future tasks, instead of writing the same brief every time.
Tasks can now be scheduled for a specific future time instead of only being placed immediately.
The phone number a call can be routed back to is now verified before use, so a callback never lands on the wrong line.
The first version of ExeclyAI is live: describe an outcome, and it plans, calls, and reports back on real phone calls.
Looking for the customer-visible history only — no internal engineering changes are listed here, and nothing about a security issue is described until it's fully resolved.
Something missing that affected you? Report a bug or contact support.
