OAKOAK/memory
Dashboard
DocumentationResources
OAK.memory

Guides

Practical paths for using memory in a personal project, a shared team store or a workflow with several agents.

Remember project decisions

Start with one private store. Save a small architectural decision with its reason and project subject. In a later conversation, recall the topic before making a related change. Update the original memory if the decision changes so the next session sees current context.

Quickstart

Build shared team context

Create a project store and give collaborators the role they need. Each person connects with their own key. Preserve explicit subjects and quotations when saving facts about several people. Check each collaborator’s credits before depending on semantic retrieval.

Connect Claude Code

Coordinate several agents

Use a dedicated store profile for each project and verify the selected store at the start of a session. Several agents on one account share request and embedding budgets. Pace requests, avoid redundant policy reads and retrieve existing memories before creating duplicates.

MCP tool reference
Documentation

Persistent memory for your AI workflow. Start with a private store, connect Claude Code, and carry useful context from one session to the next.

Quickstart

Create a store, connect your assistant and save your first memory. The cloud service runs the database for you.

Core concepts

Understand memories, stores and retrieval before connecting them to a longer workflow.

Connect ChatGPT or Claude

In your app’s custom MCP or connector settings, add https://www.openaikits.com/api/mcp and choose OAuth. Sign in to OAK, choose a store and approve access. You return to your app automatically. Automatic registration needs no API key, client ID or secret.

Connect Claude Code

Use a personal key and store profile to connect the OAK.memory plugin to hosted memory.

Share a memory store

Give collaborators access to a shared context while keeping each account’s identity and usage separate.

MCP tool reference

The cloud MCP endpoint exposes tools for store discovery, retrieval and memory changes. Your connection supplies the account and store context.

Usage tiers and limits

Understand which limits your account shares, how operation credits work and what to do when a request is refused.

Troubleshooting

Identify whether an issue comes from your connection, permissions, usage or the embedding service.

Local plugin setup

Run memory on your own machine instead of using a hosted cloud store. This has a separate database and setup process.

Guides

Practical paths for using memory in a personal project, a shared team store or a workflow with several agents.