Why an open memory protocol
Advancing decentralized, innovative, and competitive AI markets requires open standards for key components, and especially for memory. This grows the pie for everyone by establishing a single market.
Why memory matters. Memory is the persistent state an agent carries across steps and sessions: user-specific facts, past experience, and task progress. It drives personalization, agent efficacy, and the goals agents can pursue.
Shared memory. OMP is focused on transferring a core memory object, and allowing for it to be read and written to concurrently.
Today, AI memory is a walled garden:
- For end-users, it is bundled with consumer AI products, limiting functionality and shared capabilities, raising switching costs, and foreclosing competition.
- For enterprises, lock-in risks are real. Agent harnesses such as Codex or Claude Code keep data in their own formats, sometimes on their own servers, making switching products more than just switching an API provider. Provenance is often lost when memory is copied, and permissions may not survive export.
Standardizing how complementary products fit means that anyone can enter the market and innovate against the openly specified interface. Open protocols therefore help unbundle vertically integrated production, facilitating distributed innovation and, ultimately, more participatory market architectures.






