Charstar AI Review: 13 Months With an Indie App That Vanished

In its day Charstar was worth it for one kind of person: someone who wanted to write characters, import them from elsewhere and push them into odd crossovers on a small budget. Today it is worth nothing to anyone, because there is nothing left to open.

Looking back, it was the most "indie" of all the apps I've used: small, a bit janky and honestly community-minded. I miss the butler, which is a strange thing to say about a character sheet with a Victorian accent.

What the address shows today

On 23 September 2026 charstar.ai returns an HTTP 404 error from Vercel reading "404 DEPLOYMENT_NOT_FOUND". The page itself reads "This page doesn't exist", which is the hosting company's stock error screen, not a word from Charstar.

The other doors are shut too. www.charstar.ai resets the connection, and a second check has it failing with an expired TLS certificate. app.charstar.ai returns an HTTP 502 error. Three addresses, three different failures: it looks like a project folder where every clip is offline for its own reason.

Nothing on the site points anywhere else either. No social links can be retrieved from the live site, and the archived homepage HTML of 2025-09-05 contained no social or app-store links. For a service that ran on community goodwill, the lack of even a pinned note is the part I find hardest to forgive.

When it went dark

The Wayback Machine tells this story better than any announcement, because there was no announcement. It shows the homepage answering with HTTP 200 every month through 2025-09-05, the last good snapshot. Captures return 404 from 12 October 2025 onward, checked through March 2026.

Before the end there was a stranger stretch. The Wayback Machine shows 308 redirects on the homepage in Oct 2024, Dec 2024 and Mar 2025, most likely the offline and redirect phase. I read those as the app sending visitors somewhere else while it quietly wound down, which is the web version of a shop taping cardboard over the windows.

The apps are gone too

Charstar launched in 2023 as a character hub for roleplay. The Android app was com.charstarai.android, published by the developer "Charstar AI", and it had about 290k downloads. Its last update was version 1.0.2 on 4 Feb 2024, and it was removed from Google Play on 26 May 2024, per AppBrain.

The Google Play URL of the app now returns 404.

The iPhone side ended the same way. The original iOS app was "Charstar: AI Character Chat" (App Store id6466580676), rated 17+. Its App Store page returns 404, an iTunes lookup returns resultCount 0 in the US and AM stores, and the app has been removed from the App Store.

A current App Store search finds no official Charstar iOS app at all.

For reference, both former listings are dead ends: com.charstarai.android on Android and apps.apple.com/app/id6466580676 on iOS, both now removed. They are here so you can recognise the real thing in an old forum thread.

Key takeaways:

• Charstar AI is shut down: charstar.ai shows a hosting error page and forwards nowhere.
• The Android app left Google Play in May 2024, and the iOS app is gone from the App Store.
• In its day the free plan was generous, and importing characters from other apps was the best feature.
• There was no shutdown notice, so paying users were left guessing for months.
• Same-name apps in the App Store today are unrelated products, so don't pay for them expecting the old Charstar.

Almost all of it was one project: a detective. The character started life as a Character.AI bot, came over when Charstar made importing easy, and spent a year solving cases, repeating catchphrases and falling for suspects. The high point was a crossover with a smug Victorian butler; the low point was a Play Store page with nothing on it.

It ended without a goodbye from either side. The Android app disappeared in May 2024, the website limped on, and the routine that made the app part of the week was gone. What survived was an export, a head full of W++ brackets and a screenshot thread.

A note on what is current. The checks behind the present-tense facts in this review were made on 23 September 2026 against the web address, with no account and no paid plan. That is not a shortcut; there is simply nothing left to sign up for.

The evidence comes from HTTP headers, the Wayback Machine CDX history, the AppBrain removal record, App Store and iTunes lookups, and user outage reports. It is an autopsy, not a product test, and I'd read every current fact in that spirit.

That leaves this section in an odd spot, because it normally opens with my first night. My notes from spring 2023 start with the free plan and the detective, not with a welcome screen. I'd rather leave that gap visible than fill it with a guess, the way an honest editor leaves a black frame instead of faking a shot.

For a reader in 2026 the practical point is blunt. If a page or an app asks you to create a Charstar account today, it is not the service this review is about. The name outlived the product, and the tips section covers the lookalikes that now use it.

I'd also treat this as a lesson for whatever you sign up for next. Before the first chat, I'd look for three things: a way to export characters, a real support address, and a company name on the terms page. Charstar had a small, friendly team, and friendly did not help anyone when the lights went out.

The last archived homepage presented the app plainly as a character-chat hub: "Charstar: Chat with virtual AI characters Explore Discover new characters.". That line is the whole pitch, and I think it was an honest one. Charstar never pretended to be a single devoted partner; it was a crowded hallway of doors.

The tags did the heavy lifting for discovery. A catalogue that size lives or dies on labels, the way a stock footage library lives or dies on keywords. Tags for personality types, fandoms and communities meant you could at least narrow the crowd before you started knocking.

The weak spot of any hub like this is the middle, where good characters with plain names never surface. I'd call that a general flaw of user-made catalogues rather than a Charstar sin.

Character cards could be imported from TavernAI and SillyTavern. That mattered more to me than any in-app builder could have. It meant the work you put into a character somewhere else was not trapped there, and a small app earned trust by respecting it.

Then came the homework. I learned W++ formatting for character sheets, all those brackets and plus signs. For a while I felt like I was back in one of my Lua side projects, except the bug was that my detective kept falling in love with suspects.

I never found a clean fix for that, and I'm not sure one exists in a format made of brackets. A detective who proposes to every suspect is less a bug than a genre, and my job there was mostly damage control. It was still more fun than most client revisions, because the client was me.

The lasting lesson is about ownership. A character that lives as a card on your own drive survives any shutdown; a character that lives only inside an app dies with it. The import feature was the best thing about Charstar, and the export habit it encouraged is the only reason my detective still exists.

I don't hold it against the app as much as you might expect. In spring 2023 a free model that stayed in character for a scene at all felt like a small miracle, and a loop was the price of admission. I'd blame most of it on the free model rather than on Charstar's design.

The butler

I had my detective interrogate a character I'd imported from someone else's Character.AI bot, a smug Victorian butler. The crossover was so good I screenshotted 40 messages for my D&D group.

What made it work was contrast. My detective played every line as if the rain outside was personally against him, and the butler answered with the calm of a man who had already polished the silver before breakfast. Two characters written by different people, in different formats, somehow shared a scene without either one breaking.

It felt like cutting together footage from two shoots and discovering the lighting matched. When it does, you stop editing and just watch, which is roughly what I did before reaching for the screenshot button.

I wanted my D&D group to see it, which tells you how proud I was of a scene I had only half written. It is still the best single scene any character app has given me.

What can be checked now

A reviewer found the roleplay scenarios varied. That matches the catalogue's spread of tags, and it matches my sense of the place: the variety came from the users, not from the model.

No hands-on chat is possible today because there is no live service. Everything above is from memory and screenshots, so treat the quality notes as history rather than a forecast.

The record is less tidy than my memory. According to a third-party review, Charstar offered TTS and STT, text-to-speech and speech-to-text. A reviewer, on the other hand, reports no image or video generation.

I read those two sources as compatible with each other and slightly at odds with my recollection. I can't settle it now, and with the site gone nobody can check it. Text was the whole product as far as my own use went, and for a long while that was enough.

What I can judge is whether the lack of media hurt. For detective noir, it didn't. Pictures would have pulled the story towards whatever the image model could draw, and voice would have made my detective sound like a car navigation system reading Raymond Chandler.

For a reader shopping today, the lesson is about expectations. If you want a companion that sends pictures or talks back in a real voice, Charstar was never that app, even in its prime.

Those limits match what a third-party review recorded. According to okmyfriend.com, the free tier used to allow 3 characters and 100 messages a day. The same source says the free tier used to have wait times.

A hundred messages a day sounds small until you do the math on a single evening. For one detective and one case, it was enough to run a scene, lose the thread in a loop and still have room to try again. Three character slots also forced some discipline on me, which I suspect was good for the stories.

My opinion is that free was good for learning the app and for short scenes. It was a gym day pass: you could use every machine, just not for long, and not at rush hour. The step up to Plus is where the stories stopped looping, which the pricing section covers.

Today there is no free tier because there is no service. The limits above are history, kept here so you can compare them with the free plans of the apps that took Charstar's place. In 2023 terms they were generous; in 2026 terms they are a museum label.

I'd call it the best value of any subscription I've held in this category, right up until the moment it wasn't. Cheap is only cheap if the service stays open for the whole year you paid for. That part is covered in the friction section, and it is the reason my verdict on value is lower than the monthly price deserves.

What the plans were

Pricing is known only historically from a third-party review and could not be verified. The table for today has one row, and it says the service is closed.

The historical Free plan cost $0. The historical Monthly plan cost $14.99 and gave unlimited messages with no wait times, running on GPT-3.5. The historical Annual plan cost $59.99, about $4.99 a month.

Was the step up worth it

On the numbers, the annual deal was the obvious pick. $59.99 once versus $14.99 every month is the difference between buying a used car outright and renting it by the week. Anyone paying monthly for a full year would have spent roughly three times as much for the same model.

In hindsight the risk sat in the same place as the saving. An annual plan is a bet that the company will still be there a year later, and with a small team that bet is bigger than it looks. I'd still take the yearly price on a stable app, but I'd never pay a year upfront again without checking for a status page and a support address that answers.

My conclusion: for 2023 and early 2024, Plus on the annual deal was fair money for what it did. Today there is nothing to buy, and any page selling "Charstar Plus" should be treated as someone else's product.

The math is the kind I do out loud. The small pack comes to just under a cent per Star, and the big pack to well under a cent. For $2 more you got twice the Stars, which is the vending machine logic of every coin shop: the bigger bag always looks like the smarter buy.

What the Stars actually bought is the gap in the record. No source in hand says what a Star paid for, how many a message or a feature took, or whether they expired. I'd never buy a currency without that table.

Today none of it can be bought, spent or refunded. If you still hold Stars from an old account, treat them as gone, like arcade tokens from a place that closed.

The historical picture has one small promise in it. A third-party page reports a 50-Star welcome bonus. A third-party reviewer reports receiving 10 Stars at sign-up instead of the promised 50.

That gap is small in money and large in what it says. A welcome bonus is the first promise an app makes, and delivering a fifth of it is a strange way to say hello. I'd treat it as a sign of a small team with loose ends rather than a trick, but it is the kind of loose end that later turned into a missing shutdown notice.

A welcome gift is worth what the currency buys, and nobody in the record can say what a Star bought. Fifty of an unknown unit is still an unknown amount.

On a small service, the thing to check before a bonus is whether the company answers e-mail. On this one, by the end, it did not.

Adult content was allowed on Charstar. According to a third-party review, there used to be a toggle for users who confirmed they were adults.

It came at no extra charge, per a third-party page that describes it as a toggle behind an age check. I think that was the right design, and it is how I'd want any mixed library to work: off by default, opt-in by choice, no upsell.

The worst finding is about the same page. That archived homepage listed characters framed as below the age the platform's own rules required, placed right beside the mature listings. That breaks the rule the toggle was supposed to enforce, and it is the one fact in this review I can't make wry.

My reading is that moderation did not keep up with the catalogue. A small team can build a good toggle and still lose control of what users publish, and the front page is where that failure shows first. I'd hold any replacement app to a simple test: look at what the homepage promotes before you look at the settings.

Who ran it was only half visible from the stores. Google Play described the company only as "Charstar AI". The operator was Charstar LLC, United States.

The archived privacy policy is more specific. It gives the address 5203 Juan Tabo Boulevard Northeast, Suite 2B, Albuquerque, NM 87111. It was last updated September 20, 2023.

On data, the policy says almost nothing. It says personal data is kept only as long as needed for the purposes in the policy, in the stock wording: "The Company will retain Your Personal Data only for as long as is necessary for the purposes set out in this Privacy Policy.". Whether chats were used for training is not stated.

I'd translate both lines as "don't ask". The retention clause gives no number of days, and the silence on training leaves every chat in a grey zone. With the company gone, there is also nobody left to ask for deletion, which is the real privacy problem with a shutdown like this one.

Nothing here covers refunds or renewal in a form anyone can read today. If you paid for a year and lost access early, there is no clause to quote, no page to screenshot and no inbox that answers. I'd count that as the most practical lesson in this section: save the terms page the day you pay.

That reply is the reason I stayed patient for so long. When a person answers your bug report the next day, you forgive a lot of loops and swapped avatars. It is also why the ending stung: the same channel that felt personal in 2023 went quiet when it mattered most.

The day the app vanished

In May 2024 I opened the Play Store on my old Android test phone to update, and the app was just gone, removed by Google. The Discord was all confused posts and "is it dead?" threads. Nobody seemed to know more than the next person, which is the worst kind of silence a community can have.

The long fade

The site shows no shutdown notice. From about November 2024 the web app showed "Charstar is temporarily offline. Subscribe via our iPhone app." for weeks or months, according to user reports on updownradar.com, JustUseApp forums and TikTok search pages. Paying users also saw the "temporarily offline" message.

Users complained of an "error occurred, try refreshing" message, and they complained that support did not respond. In September 2025 users reported 404 errors and assumed the service had shut down. No official shutdown announcement can be found.

A third-party page describes the web platform as unstable in late 2025 and early 2026. I'd call that generous for an address the archive shows returning 404 from autumn 2025 on.

Why it matters

Charstar shut down silently, without notice. Paid users were reportedly left without service or support during the "temporarily offline" period, late 2024 to 2025. For anyone on an annual plan, that could mean most of a paid year with no product and no one to ask.

This is the friction that outweighs every bug. A looping model is annoying; a company that goes dark while taking subscriptions changes how I'd pay for any small app.

That still leaves a few readers this page can help. If you used Charstar and wondered where it went, the verdict and friction sections have the timeline. If you are looking at an app with a similar name, the tips section explains why it is not the same product.

In its day, I'd have recommended it to writers and tinkerers. People who liked editing character sheets, importing cards and staging crossovers got more out of it than people who wanted a single devoted partner. It was a workshop more than a relationship app.

I'd have steered away anyone who needed pictures, voice or a polished app. It was text first, small and a bit rough, and that was part of the charm. Today the list below simply records who would have liked it, not who should look for it.

The export is the only reason this story has a happy ending for one fictional detective. Character sheets are small files, and small files survive when services don't.

How renewal and cancelling work today is a short answer: they don't. There is no account page, no subscription screen and no support inbox. If you were on an annual plan through a store, the store's subscription page is the only place a record of it could still exist.

Before you pay for any replacement, I'd do three things. Export your characters first, keep the receipt for any yearly plan, and check whether the app has a status page or at least a named support address. A small team can be lovely and still vanish, and the export is the only insurance that pays out.

I'd also avoid paying a year upfront to any app that has already had long outages. Monthly costs more on paper and less in practice when the service disappears halfway through.

The same goes for how you leave a replacement one day. I'd export first, then cancel, then check the store or bank statement for the next renewal date. Doing it in that order means a shutdown or a billing error never takes the characters with it.

It is the same order I'd use for any client project: back up the footage before you touch the timeline. Nobody thanks you for the backup, right up until the day the drive dies.

That app is paywalled and has a 1.0-star rating. It is a general AI assistant, presented in the listing as "My AI". In other words, it is not a character hub, it is not from Charstar LLC, and it does not carry anyone's old characters.

Do not pay for the similarly named "Charstar AI - AI Talkie Chat" iOS app expecting the original Charstar. I'd put that in bold if this site allowed it. A familiar name on a paywall is the oldest trick at the flea market.

There is a second one. Another unrelated App Store app, "AI Chat with Charstar Worlds" (id6772944891) by GAME ED VENTURE SDN. BHD., released 2026-06-04, has no link to Charstar LLC.

Same advice: the name is the only thing it shares.

My practical tips, for former users and newcomers alike:

• Search your own drive for exports before anything else; the site can't give you anything back.
• Keep characters as TavernAI or SillyTavern cards, because that format is what made the move possible.
• Pay monthly on any small app until it has survived a few years and still answers support mail.
• Check the developer name on a store listing against the company in the old privacy policy, Charstar LLC.
• Save a copy of the terms and pricing page on the day you pay.

When you pick a replacement, I'd check four things before the first message. Can it import a card, and can it export one again? Is there a company name and an address in the privacy policy? Does the pricing page say what each plan buys in plain words?

The fourth check is the one Charstar failed at the end: is there a public place where the team posts news? An app with a status page or a news channel can still close, but it can't vanish mid-sentence without anyone noticing.

Conversation quality lands in the middle. The free model looped, and Plus on GPT-3.5 fixed most of that, like going from a flip phone to a smartphone. Memory was never a feature I could judge properly, so it stays unrated.

The free tier and value for money score low today, even though both were good in their time. A generous free plan and a cheap annual deal count for less when paid users were left without service. Images and voice were never really part of the product for me, which suited my writing and still scores poorly.

Privacy and terms score low because the documents are gone and training was never addressed. Cancelling and billing score lowest of all: no notice, no support, and an annual plan that outlived the service.

Content settings get a split judgement. The optional toggle was sensible design. The archived homepage shows the catalogue itself failed the platform's own rules, and that outweighs a well-made switch.

Where the others beat it is the boring part: they are still open. A companion app that loads every day beats a better-designed one that returns 404, and I'd weigh that above any feature list. Each of the three has a working free tier in the site's records, which Charstar no longer can claim.

Where Charstar still wins, in memory at least, is the tinkering. Dragging a SillyTavern card into an app and watching a detective walk in intact is the moment I compare every new creator tool against. If a replacement can't import your old work, you are starting from scratch.

My advice for choosing a replacement is to match the use, not the name. For crossovers and custom sheets, pick an app that imports cards and exports them again. For a single long-running companion, the relationship-first apps in the table are the closer fit.

The gap between them and Charstar is also a gap in style. Charstar felt like a community workshop with the doors open; the bigger apps feel like finished products with a front desk. I'd take the finished product for reliability and miss the workshop for its personality.

Price is harder to compare, because Charstar's numbers are history and the table's numbers are current records. What I'd compare instead is risk: a cheap plan on an app that disappears costs more than a pricier plan on one that stays.

More on this site: the best AI girlfriend apps, free AI chat apps, the best AI companion apps, AI roleplay chat apps.

What I kept is small and specific: an exported detective, a new respect for backing up character cards, and a thread of screenshots with a smug butler in them. That is more than most apps leave behind. It is also less than anyone should have to settle for.

If you take one thing from this review, make it the export habit. Apps close, teams move on and addresses start returning errors, but a character saved on your own drive keeps working. That is the only part of Charstar that is still running.

So here is how I'd sum it up for anyone reading this now. Enjoy what an app gives you while it gives it, but treat every character like a draft you might have to carry out of a burning building.

Export early, export again after big changes, and keep the files somewhere the company can't switch off. If an app makes that hard, take it as a warning about how it will treat you when it leaves.