The OWE concept and ASCP only works if it becomes part of the fabric of computing, not a feature of a single product. That’s why ASCP is being developed as an open, international standard—a protocol that anyone can implement, extend, or build on.
Modern AI systems suffer from the same fragmentation that has plagued the rest of computing: every tool, app, and model holds its own isolated memory. This creates a world where context can’t flow, collaboration can’t compound, and agency is slowly ceded to platforms. Fixing this doesn’t mean building a bigger model or a better app. It requires shared infrastructure.
ASCP is designed to be that infrastructure. It is:
-
Open and vendor-neutral
No company should own the world’s collaborative memory. ASCP is structured like other public protocols—HTTP, DNS, Git—so the ecosystem can remain interoperable, durable, and resistant to capture.
-
FOSS-powered and community-driven
The reference implementation is planned as open-source software so builders, researchers, and integrators can inspect, extend, and test it. ASCP grows through real adoption, not locked-down APIs.
-
Internationally governable
The protocol is being written toward IETF-style standardization so multiple stakeholders—developers, researchers, institutions, companies—can steward its evolution. Long-term durability requires global oversight, not reliance on a single platform.
-
Ecosystem-first rather than product-first
Reframe’s implementations (like the AI Desktop) are examples meant to inspire and accelerate the community, not to enclose it. Our goal is a world where many clients, agents, models, and tools speak ASCP natively.
-
A foundation for shared cognition across tools and agents
The real opportunity isn’t a better chat interface; it’s enabling durable collaboration across human teams, autonomous agents, and diverse applications—without forcing everything into one ecosystem.
Our long-term objective is simple:
ASCP becomes the protocol layer that the next generation of tools, agents, and operating environments rely on for shared context.
To get there, we’re working with:
who see that the future of AI cannot be a collection of walled gardens.
By grounding the OWE and Agents in an open protocol, we enable a computing world where context moves freely, agency stays with the user, and collaboration compounds over time—regardless of which model, tool, or platform you choose.