Kyrelo is a desktop application, not a browser extension and not a cloud dashboard, built to replace two very different categories of paid social media tool with one free one. On the publishing side it schedules posts to X and to Bluesky, checks each platform's character limit as the user types, and can hand off an entire content plan to an AI model that researches a product or niche and spreads a series of posts out at natural, uneven times. On the cleanup side it does what TweetDelete and its many clones have charged for over the years: bulk-deleting old tweets and reposts, and unfollowing accounts that have gone quiet or look automated. It is, in effect, Buffer's scheduling calendar and TweetDelete's mass-delete button fused into one program, minus the invoice. The technical detail that matters most is how it connects to X. Kyrelo does not route requests through X's paid developer API at all. Instead, when a user connects their account, Kyrelo opens a real Chrome browser window on their own machine, the user logs in exactly as they normally would, including Google or Apple single sign-on, and that browser session is saved locally. From then on, the app clicks and scrolls through that session the way a person would, rather than making metered calls to X's servers. That single architectural choice sidesteps the API pricing and access tiers that forced many third-party Twitter clients out of business after the platform tightened developer access in 2023, most visibly when long-running apps like Tweetbot and Twitterific lost access overnight. It also explains the throttling built into the unfollow tool: a few seconds between each action, a cap of 100 unfollows per run and 300 per day, and an automatic stop if X starts refusing requests, all designed to look like a cautious human rather than a script hammering an endpoint. The project is released under the MIT license, one of the most permissive open-source licenses available, meaning anyone can read the code, modify it, fork it or bundle it into their own product with almost no restriction. It carries the GitHub handle sobytes as its maintainer, and the site frames the entire project as a reaction to relying on someone else's infrastructure: a section on the marketing page points directly at Buffer's own public status history, describing multi-hour outages across web, iOS, Android and the API layer and citing a rough uptime figure of about 97 percent over a recent quarter, according to Kyrelo's own site. The pitch, in the developer's words on that page, is blunt: cloud schedulers go down, and Kyrelo runs entirely on the user's machine, so if something breaks, it breaks for one person alone instead of every customer at once. There is no subscription, no account creation and, as far as the published material shows, no built-in monetization at all. There is no Kyrelo server to bill for, no support tier to upsell and no ad-supported free plan gating features behind a paywall. That is the whole appeal, and it rests on a genuinely new assumption: that a single maintainer, working with AI coding tools rather than a team of engineers, can keep pace with the kind of platform changes, a tweaked login flow on X, an API update on Bluesky, a code-signing requirement from Apple or Microsoft, that used to require a company's worth of staff to track. Right now that maintenance runs on one person's spare time, aided by the same kind of Claude and OpenAI models the app itself offers to users, a bet that AI-assisted, community-driven upkeep can hold up on its own without a bigger organization standing behind it. Set against Buffer and TweetDelete feature by feature, the trade-offs are stark in both directions. Buffer supports a long list of networks beyond X, including Instagram, LinkedIn, TikTok, Pinterest, Threads and YouTube, offers analytics dashboards, team seats, client approval workflows and official mobile apps, and keeps posting even if a user's own laptop is off, asleep or on a different continent, because the scheduling happens on Buffer's servers rather than the user's hardware. TweetDelete similarly runs its mass-deletion jobs in the cloud, independent of whether the user's computer is switched on. Kyrelo cannot match any of that: it only speaks to X and Bluesky, has no analytics or team features, comes with no formal customer support desk, leaning instead on its open-source community and, increasingly, on the same AI assistants users can turn to for troubleshooting, and, critically, a scheduled post will not fire if the machine running Kyrelo is powered down when the clock strikes the chosen time. The iPhone companion app included with the project does not change that; it is a client that connects to the desktop instance over local Wi-Fi or Tailscale, a private networking tool, so the phone can trigger and monitor jobs, but the actual work still happens on the computer sitting at home or in an office. On adoption, there is little that can be stated with confidence right now. The project's public presence centers on its own website and GitHub repository rather than any third-party coverage, and concrete figures such as download counts, star totals or independent user reviews were not available in the material reviewed for this piece, so none are repeated here as fact. Kyrelo is not alone in trying to build a self-hosted, non-subscription alternative to social media schedulers; open-source projects such as Postiz and Mixpost occupy a similar lane, aiming to let people run their own posting infrastructure rather than rent someone else's. Whether Kyrelo carves out real usage against that competition, or against Buffer's much larger feature set, is an open question with no verified answer yet. The privacy pitch is where Kyrelo is most convincing on paper. Because there is no Kyrelo account and no Kyrelo backend, login sessions, any Claude or OpenAI key added for the AI features, and Bluesky's app password all live in a local data folder on the user's own device rather than on a company's servers somewhere else. The security trade is genuine but double-edged: nobody else can be breached and leak a user's X session, but nobody else is auditing that local folder either, and anyone running the app is trusting a small, largely single-maintainer open-source codebase with the keys to their accounts. Deleted posts, the site is explicit about this, cannot be recovered once removed, so the same lack of a safety net that makes the tool cheap also makes mistakes permanent.
🌍 world
A Free, Local Alternative to Buffer's Subscription Trap
A new open-source app called Kyrelo wants to do what Buffer and TweetDelete do, scheduling posts and mass-deleting old tweets, without a monthly bill or a company's servers standing between a user and their own account. It runs entirely on a Windows or Mac machine, logs into X through a real Chrome window instead of a paid API, and stores every credential in a folder the user controls.
My Take
Give credit where it's due: the core insight here is correct. Buffer and its rivals sell convenience wrapped around infrastructure that regularly falls over, and paying a recurring fee for the privilege of trusting somebody else's uptime has always been a strange trade for something as simple as queuing a tweet. Building that scheduler as a local app that a user actually controls, with the login session and API keys sitting in a folder on their own hard drive rather than a stranger's cloud, is the right instinct, and the browser-automation trick for dodging X's punishing developer API pricing is genuinely clever engineering. But the old assumption that a project built by one person in their spare time is automatically fragile no longer holds the way it used to. AI coding assistants like Claude and OpenAI's models mean a single maintainer today can patch a broken login flow, adapt to a platform's API change or ship a new feature at a pace that once required a dedicated engineering team, not a lone hobbyist working through the problem alone. Kyrelo is itself a small proof of that shift: the project already leans on Claude and OpenAI to help draft users' content plans, and there is no reason the same kind of tools can't be turned inward, on the app's own codebase, whenever X or Bluesky change how they work. That also changes who can safely run a tool like this. It used to be a reasonable bet only for a technically comfortable user who could read code before running it, and a poor one for anyone who didn't know what MIT licensing or a Tailscale network meant. That gap is closing: an AI assistant can now explain in plain language what a permissive open-source license actually entails, walk a non-technical user through setting up a Bluesky app password, or flag when a piece of software is asking for access it shouldn't need. The steep learning curve that used to gate open-source tools to developers alone is flattening fast, and that changes the risk calculation for everyday users, not just power users. The bigger shift is structural, not just technical. The old logic said a service like this needed a company behind it: a support desk, a product team, a payroll big enough to keep the lights on when something broke at 2 a.m. Kyrelo, and a growing wave of community-run, self-hosted projects like it, are testing whether that assumption still holds when a single motivated maintainer, working with AI tools and an MIT license that invites anyone to fork or contribute, can do the job just as reliably as a larger organization once needed to. It won't look like a SaaS (Software as a Service) company, with its uptime guarantees and its subscription invoice, and it may not need to. Enjoy Kyrelo for what it is today: an early, sharp-edged example of a model that increasingly does not require a large company standing behind it to hold up over time.
What Happens Next
The immediate test for Kyrelo is whether X Corp's live site keeps tolerating the kind of browser automation the app depends on for login, posting, deleting and unfollowing. Platforms have historically clamped down hard on tools that behave like scripted browsers rather than paying customers of an official API, and any future detection change on X's side could break Kyrelo's core functions overnight, leaving the fix in the hands of whoever maintains the sobytes repository on GitHub. Expect scrutiny to build around sustainability rather than features next, though the terms of that scrutiny are shifting. A tool with no revenue and one visible maintainer used to be assumed fragile by default; increasingly, AI coding assistants let a single maintainer cover ground that once needed a team, and an MIT license leaves the door open for outside contributors to fork or extend the project if the original author's attention moves elsewhere. Watch the GitHub repository's commit history and issue tracker over the coming months for the real signal: active pull requests from outside contributors, or evidence that fixes are landing quickly with the help of AI-assisted tooling, would both suggest staying power, while a slowing commit log with no community pickup would suggest the opposite. If the model does prove durable, the more interesting long-term question is platform coverage. Kyrelo currently only touches X and Bluesky; adding Threads, Mastodon or LinkedIn would put real pressure on Buffer's pitch for casual and solo users, though it would also mean maintaining browser automation against several moving targets at once rather than one, a task that is far more feasible for a small team now than it would have been a few years ago. Competing self-hosted projects such as Postiz and Mixpost will be worth watching too, since a crowded field of free alternatives could either validate the whole category or split it into tools too thin on users to keep improving.
What History Tells Us
Kyrelo's bet on browser automation instead of an official API echoes a fight that has played out before. In January 2023, X's predecessor Twitter abruptly cut off access for long-established third-party clients including Tweetbot and Twitterific, part of a broader move that priced most independent developers out of its official API entirely. Tools built since then have largely had two choices: pay for increasingly expensive official access tiers, or automate a real browser session the way Kyrelo does, accepting the risk that the platform could target that behaviour directly at any time.