TL;DR
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
Exact is a new, pre-1.0 cross-platform application runtime described in its GitHub repository. It uses a shared interface language and supports web, macOS, iOS and Linux, with the repository emphasizing native platform rendering and agent-driven testing. Its developers warn that the project changes daily and has no API stability or migration guide.
The developers of Exact have presented a new cross-platform application runtime designed to run one interface on the web, macOS, iOS and Linux. The project combines a declarative interface language called Contract with TypeScript or Rust for application logic, and says native versions use platform-specific rendering rather than a webview. Exact is pre-1.0, and its maintainers warn that interfaces and APIs may change without compatibility guarantees.
According to the GitHub repository, developers describe an interface in Contract, a small declarative language, and write data logic in TypeScript or Rust. A compiler turns the interface into a plan. On the web, a compact runtime uses that plan to drive the browser’s DOM. On macOS, iOS and Linux, a Rust runner executes it, while a Rust layout kernel calculates positions and each platform draws the interface with its own tools: AppKit, UIKit or a GPU painter. The project says there is no webview or JavaScript bridge in the native UI, and that app JavaScript does not run before the first pixel.
The repository’s example workflow asks a coding agent to generate a to-do app, add a Contract test and run it on the web, macOS and the iOS Simulator. It provides a shared agent driver for operations such as taking screenshots, tapping controls, entering text and reading app state. Tests written in app.test.contract are intended to run unchanged on supported hosts. The project says the driver can also target a real iPhone and Linux, though the supplied demonstration focuses on web, macOS and iOS.
The repository reports that an agent completed an earlier version of the example on a clean clone in about 57 minutes on October 4, 2026, with tests passing on all three platforms. About 25 minutes went to initial macOS and iOS builds, according to the report; it says later builds take a minute or two. These are project-reported results from a specific setup, not an independent comparison. The repository also lists performance targets and benchmark results, including a 99-millisecond edit-to-page update on a fresh app, but the material does not provide independent verification.
One Interface Across Native Hosts
Exact’s pitch addresses a familiar trade-off in cross-platform development: sharing application structure while retaining interfaces that use each operating system’s native controls and conventions. If the project’s approach works as described, teams could reuse interface definitions and tests without delivering a web page inside a native shell. That could matter to developers who want one codebase to cover browser and desktop or mobile users.
The project also treats AI-agent testing as a core part of its design. Its driver gives an agent a common way to inspect and operate an app across platforms, while Contract tests can be reused between hosts. This may make automated development workflows easier to reproduce, but the repository’s example does not establish how reliably agents will build production applications or how the framework performs across a broad range of devices and use cases.
For prospective adopters, the project’s maturity is a direct consideration. Exact says it has no API stability or backward compatibility. That makes it an experimental option rather than a settled foundation for applications that need predictable upgrades.
Products Worth Considering

Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite

Apple AirPods Pro (2nd Generation) Wireless Earbuds with USB‑C Charging, Up to 2X More Active Noise Cancelling, Transparency Mode, Adaptive

DJI Mini 3 Drone with DJI RC and 3-Axis Mechanical Gimbal
As an affiliate, we earn on qualifying purchases.
A Rebuild With Early-Stage Limits
The repository identifies this project as exact2, a rebuild of an earlier version called exact1. It says design documents imported from the predecessor are research material rather than authoritative specifications. The maintainers’ stated principles include designing for AI agents, tracking performance budgets and using web development concepts such as CSS while leaving platform controls and behaviors to the host system.
The project explicitly describes itself as pre-1.0 and changing daily. It says there is no migration guide by design and points users to its deferred-rules document. An October 16, 2026, upgrade note describes a change to how Snapback mutation failures are returned. That note illustrates that implementation details can shift, but does not amount to a general compatibility policy beyond the project’s stated warning.
The repository also features Softwear, a demonstration store generated from exact-templates. It uses Alternative Apparel’s catalog, but the project says it is not affiliated with that brand and does not accept real orders or payments. The demo is evidence of an example application, not a commercial integration or proof of production readiness.
“Write an app once. It runs natively on the web, macOS, iOS, and Linux.”
— Exact project repository
Performance and Adoption Still Unverified
The repository provides its own implementation details, timing figures and benchmark claims, but the supplied material does not include an independent review, broad third-party testing or comparative methodology sufficient to validate them. It is also unclear how complete support is across all four named platforms, which devices and operating-system versions are covered, and whether the same features are available on each host.
Exact’s maintainers say APIs may change and backward compatibility is not provided. The repository does not establish when the project might reach a stable release, what support commitments it may offer or how existing Exact apps would be handled through future changes. The agent example documents a particular run and machine setup; it does not show how the workflow performs across different hardware, projects or coding agents.
Watch for Stability and Platform Evidence
Developers interested in trying Exact can follow the repository’s quick-start instructions and inspect its current platform support, setup requirements and deferred rules before building on it. The maintainers have not announced a stable-release date in the supplied material, so the next useful milestones would be clearer compatibility guidance, broader platform documentation and independently reproducible performance results.
Until then, the project’s own warning remains the clearest guidance for teams: Exact is an actively changing pre-1.0 framework. Any evaluation should account for the possibility of API changes and avoid treating the repository’s reported timings or agent demonstration as a guarantee for other applications.
Key Questions
What is Exact?
Exact is a cross-platform application runtime that uses Contract for interface descriptions and TypeScript or Rust for data logic. Its repository says apps can target the web, macOS, iOS and Linux.
Does Exact use a webview for native apps?
The project says its native UI does not use a webview or JavaScript bridge. It describes native rendering through AppKit on macOS, UIKit on iOS and a GPU painter for Linux.
Is Exact ready for production?
The maintainers call Exact pre-1.0 and say it changes daily, with no API stability or backward compatibility. The supplied information does not establish production readiness.
Can the same tests run on multiple platforms?
The repository says app.test.contract tests run unchanged across hosts using its agent driver. It reports passing tests for a sample app on web, macOS and the iOS Simulator in one documented run.
Are Exact’s performance figures independently verified?
The repository reports benchmark and build timings, including a 99-millisecond edit-to-page update for a fresh app. The supplied material provides no independent verification, so those figures should be treated as project claims.
Source: github.com
