Give your agents knowledge and memory.

A reserved service for application data, retrieval and agent memory. Origon runs the datastore; your team builds and runs the application.

AI Datastore combines storage, retrieval and memory, with S3-compatible APIs and native MCP.

Request Cloud Access

Keep source files and work history together.

Store application data alongside the artifacts agents produce, session records and the history of where information came from.

Find the information behind an answer.

Search by meaning, matching text and relationships. Knowledge extraction makes content available to agents, and their datastore actions are logged.

Meaning

HNSW vector search finds relevant context through similarity.

Words

BM25 full-text retrieval finds content through matching text.

Relationships

The knowledge graph connects information through its relationships.

Connect through S3 and MCP.

Control application access with identity and permissions. PII detection and redaction help handle sensitive content.

  • S3-compatible API

    Store and access objects and artifacts.

  • Native MCP endpoint

    Manage memory, sessions, and lineage.

Carry context into the next session.

Agents can retrieve earlier decisions, commitments and lessons across sessions and months. Retention settings and authorized deletion govern what stays available. Measure the effect of memory against a baseline; longer use does not guarantee better results.

Correction
Correct stored memory when information changes.
Retention
Apply configured retention to stored memory.
Deletion
Remove memory through authorized deletion.

How the datastore is built.

Objects, search, graph relationships and agent state run together in one engine, without external database dependencies.

Explore Platform Architecture

  • One engine

    Built in Rust, with objects, retrieval, graph relationships and agent state in one process.

  • Storage redundancy

    Erasure coding adds redundancy to stored data.

  • Connected datastore nodes

    Datastore nodes communicate over QUIC.

  • Replicated data

    Data is replicated across the datastore cluster.

See how it works with your data.

  • Test representative work

    Check retrieval with representative files and queries. Compare the same task with and without retained memory.

  • Plan the migration

    Identify the data, state and workflows to move. S3 and MCP compatibility still requires coverage checks and migration work.

  • Define memory requirements

    Establish access, correction, retention and deletion requirements.

  • Check your operations

    Verify the S3 and MCP operations and permissions your application needs.

Explore AI Cloud.

See AI Datastore in action.

Start with a demonstration. Then review capacity, access, retention and timing for your application.

Request Cloud Access