Yesterday's Top Launches: 2 Tools from June 26, 2026

Yesterday brought two distinct tools to the table for developers, each tackling a very different but equally frustrating aspect of building modern software. One aims to solve the puzzle of finding what you need in a sprawling ecosystem, while the other rethinks a fundamental user interaction from the ground up. If your week involves wrangling integrations or crafting user interfaces, these new developer tools are worth a look.

Stripe.Directory

For anyone who has ever spent an afternoon bouncing between Stripe’s documentation, a marketplace, and a GitHub repo just to figure out if a service exists and how to connect to it, Stripe.Directory feels like a direct answer to that particular sigh of frustration. It’s essentially a unified search engine for the entire Stripe business network.

The problem it solves is one of fragmentation. Stripe’s ecosystem has grown to include not just its core payments APIs, but a whole constellation of Stripe Apps, service providers from Projects.dev, and machine-payable services via mpp.dev. Finding the right piece—be it a specific database provider, a tax calculation service, or a fulfillment partner—meant knowing where to look. This directory consolidates all of that into a single, searchable layer. It’s less about introducing new services and more about making the existing ones dramatically easier to discover and, crucially, integrate.

What makes it interesting beyond a simple list is its dual focus on human developers and AI agents. The search functionality is available through a CLI (stripe directory search), outputting either human-friendly results or structured JSON for programmatic use. But the more forward-thinking angle is the built-in “agent skills.” The idea is that an AI agent, tasked with building a function that needs a database, could autonomously query the directory, evaluate options like Neon or Supabase, and initiate the integration steps. This pushes it from being a reference tool to being a foundational component for what Stripe is calling “agentic commerce.”

The immediate benefit is pretty clear: less time searching, more time building. A developer can theoretically find and connect a service in a more streamlined flow. For teams experimenting with AI-driven development, it provides a structured pathway for agents to operate within the Stripe universe.

A quiet observation, though, is that its ultimate value is intrinsically tied to the depth and quality of the Stripe ecosystem itself. It’s a powerful map, but the map needs a rich territory to be useful. For developers already invested in Stripe, this is a no-brainer utility that should save headaches. For those on the fence, it’s a compelling argument for the cohesiveness of that ecosystem.

FUTO Swipe

In a different corner of the development world, FUTO Swipe tackles a problem that’s ubiquitous but often taken for granted: swipe typing. Most of us use it on our phones, powered by the closed-source, cloud-dependent algorithms of Google or Apple. FUTO’s approach is to tear down that walled garden and offer the core technology as open-source, on-device models.

The challenge here is accuracy and privacy. Achieving smooth, predictive swipe typing typically requires massive datasets and cloud processing, which means your keystroke paths are often sent off your device. FUTO Swipe aims to deliver comparable accuracy while keeping everything local. This isn’t a full keyboard app; it’s a collection of small, efficient models that developers can bake into their own projects.

The architecture is clever, built around three modular components. A layout-agnostic encoder processes the raw swipe gesture. A layout-specific decoder (they’ve launched with QWERTY) translates that into probable words. A tiny context language model helps with prediction. This modularity means adding a new keyboard layout, like AZERTY or DVORAK, could be a matter of training a new decoder without overhauling the whole system. Perhaps just as significant as the models is their simultaneous release of a training dataset of one million swipe gestures—a rare and valuable resource for researchers and developers in this niche.

The benefits are all about control and privacy. For developers building custom keyboards for Android, alternative mobile OSs, or even VR/AR interfaces where traditional typing falls apart, this provides a viable, high-quality swipe engine without licensing fees or privacy compromises. For end-users, it promises a familiar typing experience in apps or platforms where Google’s Gboard or Apple’s keyboard simply aren’t an option.

An honest take is that this is a highly specialized tool. The average app developer will likely still reach for a system keyboard. But for those in the realms of open-source mobile development, accessibility tech, embedded systems, or privacy-focused applications, this is a minor revelation. It democratizes a piece of tech that has been firmly under the control of major platforms. The fact that it’s designed to run efficiently on-device opens up possibilities on lower-power hardware, too.

Yesterday’s Launches in Brief:

For a deeper dive into either of these projects, you can find more details here: