A neutral intelligence layer that describes software by capability, trust, and execution — legible to humans and to the systems acting on their behalf.
From the team behind Uptodown. Two decades of real software distribution, becoming infrastructure for what comes next.
Code has become cheaper to produce. Operational knowledge of software has become more valuable.
Every day, more software is produced than any individual, organization, or platform can meaningfully evaluate. Ecosystems have fragmented across operating systems, package managers, stores, runtimes, and distribution models — each with its own metadata, its own trust assumptions, and its own view of what a given piece of software actually is.
Discovery, historically the core problem of app stores, has ceased to be the hard problem. The hard problems are now: what does this software actually do; can it be trusted; how is it obtained, installed, updated, and rolled back — and under what conditions does any of that continue to hold.
App stores were never designed to answer these questions neutrally. They were designed to sell. As more software becomes consumed and orchestrated by agents and automated systems, the need for a neutral, structured, machine-readable understanding of software becomes infrastructural — not optional.
Atlas is not a store, not a catalog, and not an interface over a catalog. It is a foundational layer that describes software in a structured, neutral, operationally useful way — legible to humans and to the systems built around them.
What software does.
Whether it can be trusted.
How it runs.
Who consumes it.
Atlas extends work that has been in motion, in production, at scale, for more than two decades. The team behind it has spent twenty years operating real-world software distribution — indexing applications, curating metadata, running security workflows, maintaining version histories, and shipping software into every corner of the world.
This is not a thesis in search of a dataset. It is a dataset, an editorial practice, and an operation — in search of a more strategic expression.
The foundation is already there. Atlas is its public form.
Atlas is developing as a single coherent system. Its first public surfaces are focused, opinionated, and intended to support very different kinds of consumers of software intelligence.
Structured software identity at scale: canonical identifiers, capability classification, version history, cross-platform linkage, and enriched metadata that treats each piece of software as a continuous entity rather than a static listing.
Security signals, provenance, and classification designed to be legible to both humans and systems. A deliberate separation between real risk, policy conflict, and uncertainty — with the reasoning behind each signal made available, not hidden behind a verdict.
How software is obtained, installed, updated, rolled back, and operated across environments. Platform-aware compatibility and continuity information that reflects how software actually behaves — not only how it is published.
Machine-readable interfaces for third-party stores, enterprise governance, orchestration tooling, and emerging agent-based workflows. Software intelligence should be a first-class input into other systems, not a page to scrape.
Mirrors, isolated deployments, and policy-ready software intelligence for regulated environments, national digital programs, and sovereignty-oriented organizations that require independence, auditability, and local control.
Atlas is defined as much by what it refuses to become as by what it builds. These five commitments are the durable constraints underneath every design decision.
Independent of any single platform, vendor, or store. Atlas describes what is — it does not prefer who ships it.
Explainability over opacity. Provenance over assertion. Continuity over snapshots — and the reasoning behind every signal, made available.
Open access to software knowledge across regions, systems, and use cases. The fragmentation of software must not mean the fragmentation of understanding.
Software organized into operationally useful knowledge — not just lists, not just search, not just prose. Structure that other systems can act on.
Editorial judgment and human context matter more in a noisy software world, not less. Atlas is built by and with people, deliberately.
A neutral, enriched software intelligence layer beneath app stores, marketplaces, and distribution platforms — so they can focus on their own front ends without rebuilding the index from scratch.
A foundation for software inventory, approval workflows, and risk evaluation that reflects reality across operating systems, clouds, and vendors — rather than the view of a single platform.
Deployable, auditable software intelligence for regulated environments, national digital programs, and public-sector procurement where independence and local control are non-negotiable.
APIs and machine-readable context for automation tooling, build systems, deployment platforms, and agent workflows that need to reason about software rather than guess at it.
Uptodown continues as a consumer-facing distribution product, with the reach it already has. Atlas is the deeper platform — the intelligence layer that Uptodown is being rebuilt on top of, and that many other products, services, and organizations will be able to build on top of as well.
This is an expansion upward, not a pivot away. Uptodown gains a stronger foundation. Everything built above that foundation becomes possible.
Atlas is being built deliberately, not launched. Its first services will emerge over the coming quarters and years, some quietly, some as public releases, each shaped by real operational use rather than announcement cycles.
This page exists now so that the direction is legible — to collaborators, to future partners, to the teams and organizations who will build on top of Atlas, and to the small number of people who will help shape it.