Omni Logo
Developer Platform

Build private memory systems
that work everywhere.

An open-source runtime and API gateway for secure, local-first agentic memory. Spin up isolated user enclaves, manage access credentials, and compile prompt-ready contexts with zero data leaks.

Secure agentic memory. Simplified.

An enterprise-ready architectural stack built specifically for privacy-sensitive AI applications.

Gateway API

Ingest facts and build prompt-ready contexts using clean endpoints. Explicit filters coordinate scopes across users, agents, and sessions.

  • Scoped user/agent keys
  • Standard completions contracts

Enclave Isolation

Ditch shared database tables. We provision separate, dedicated Cloudflare R2 object stores and Vectorize indices for each user space.

  • Isolated Vectorize namespaces
  • Dedicated user R2 buckets

Local-first Sync

Offload embedding calculations to client devices. Blur-hashed similarity coordinates sync in the background so plain text never leaks.

  • Client-side embedding models
  • SQLite local database fallback

Choose your path

Whether you want to self-host or use managed infrastructure, Omni has a path for you.

Core Runtime

The open-source backend that powers Omni is live. Run it yourself to own your database, sync boundary, and retrieval layer.

  • Published open-source backend
  • Self-hostable runtime
  • Powers PKMS clients
View core-runtime

Omni Local

The local mobile PKMS client is published and talks to core-runtime for private capture and recall on your own terms.

  • Published local mobile client
  • Connects to core-runtime
  • Built for personal knowledge
View omni-local

Managed Cloud

Managed sync, search, and AI infrastructure is the next path for people who do not want to run their own node.

  • Managed signup in development
  • AWS and Cloudflare infrastructure
  • Hosted sync and AI roadmap

Retrieval built for agents

Auto-vectorization

Everything is prepared for semantic recall, beginning with the open-source runtime and local-first data layer.

Semantic search

Ask natural questions and find exactly what you were thinking of, even if you forgot the exact keywords.

Ready to integrate?

Get started by reviewing our API specifications or authenticating with your SSO broker.