OAK.memory in ChatGPT and Claude
Connect OAK.memory to ChatGPT and Claude to carry useful memories across conversations. Here is how to get started and search several stores at once.
Read the updateMemory as infrastructure
for AI.
Connect memories. Recall the context that matters. Associative memory infrastructure for users, projects, and agents.
ombre éphémère gravée à jamais
Connect your AI through the cloud, or build with the open-source engine.
OAK.memory · The associative memory engine
Memory that connects, recalls, and evolves.
Meaning in every fact. Connections between memories. OAK.memory combines vector search with graph traversal and manages the full memory lifecycle: store, recall, update, merge, and delete.
01
Recall connected context
Vector search finds entry points. Graph traversal expands along connections. Recall uses both semantic similarity and relationships.
02
Manage memory as it evolves
Add facts, update changes, and merge duplicates. Manage memory from creation through deletion in one lifecycle.
03
Entity-specific networks
Organize memory by user, project, or agent. Manage facts and relationships within each network.
In use · Claude Code
Recorded plugin session · See memory storage, recall, updates, merging, and deletion in action.
recall
01Find relevant memories with vector search, then follow structural links to recall connected context.
›memory.recall("what are they building?")
building OAK0.88
solo founder0.72
01 vector entry02 graph traverse
insert
02A new fact is embedded and linked into the network — instantly associated with everything it relates to.
›memory.insert("just learned Rust")
learned Rustnew
01 embed02 link
update
03Update an existing memory as facts change, while managing its connections and change history.
›memory.update("prefers dark UI")
prefers OLED darkv2
prefers dark UIold
01 match02 rewrite
merge
04Consolidate duplicate memories and remap their related links and history.
›memory.merge("ships fast" ⋈ "moves quick")
ships fast2 → 1
01 detect dup02 fuse
delete
05A memory is pruned and its edges retract — forgotten cleanly, with no dangling links left behind.
›memory.forget("ships fast")
ships fastpruned
01 locate02 prune
Two-phase retrieval
Edges carry no embedding.
A link is a pure structural connection, so what it means is interpreted at query time rather than frozen when it was created.
recall.ts
import { Memory } from "@openaikits/memory"const memory = await Memory.initialize({ db, embeddings })// vector entry → graph traversalconst ctx = await memory.recall("what does the user prefer?")await memory.remember("prefers dark, ships fast")▍Hybrid vector + graph
One unified store: embeddings for discovery, edges for association. Two-phase recall in a single call.
Entity-specific networks
Every user, persona, or agent keeps its own associative network. Personalization lives in the graph, not in the model weights.
Local embeddings
Generate embeddings locally with Ollama. Data location and external API use depend on your database and embedding configuration.
Framework-agnostic
Pluggable database and embedding adapters. Integrate the memory engine into your application and agent stack.
One engine, many surfaces
One engine. Multiple ways to begin.
Connect a cloud store to a supported AI, configure the engine locally, or integrate it into your application. Different entry points, the same associative memory foundation.
GET STARTED
Which AI do you use?
Install the published OAK.memory plugin in ChatGPT, use the Claude Code plugin or follow the connection guide for your app. Then sign in and verify your store.
Start cloud memory with free credits. Check your account for top-up and Base subscription availability, current credits and limits. Your AI provider may require its own plan.
OAK.memory plugins are available for ChatGPT and Claude. Cloud connection steps differ by app. Choose yours below.
Claude Code (CLI)
Install the plugin from GitHub and configure your cloud store profile. No local database is needed for cloud mode.
View connection guideClaude
Add OAK.memory Cloud, sign in to OAK, and choose your stores. Organization connection policies apply to company accounts.
View connection guideCodex
Connect in the desktop app or CLI and grant access.
View connection guideChatGPT
Install the published OAK.memory plugin, sign in to OAK and choose your store.
Install OAK.memory in ChatGPTManaged cloud memory
Connect an OAK store to supported AI without a local database. Sign in, select a store, and verify the connection.
Free tier and limitsLocal setup · App integration
Choose database and embedding adapters for your own setup. Data processing and external API use depend on that configuration.
Copy this prompt into the AI you use
Your AI will choose one setup path for your environment. You still need to complete any installation and consent in the app. Expand the full prompt to review its contents before copying.
I want to connect OAK.memory to my AI. Help me:
View setup prompt
I want to connect OAK.memory to my AI. Help me: 1. Tell me which app or environment you are running in (Claude web, Desktop, Code, ChatGPT, Codex, etc.). If you cannot tell, ask me. 2. Read https://www.openaikits.com/setup.md and choose the easiest single path for my environment. 3. Guide me one click at a time, without a terminal. If a terminal is required, explain why and ask before proceeding. Explain any step you can perform yourself before doing it. 4. Verify the connection with real tools and show my store name, role, separately verified OAuth scopes/tool availability, and installed version when available. If already installed, check whether it is current. Only report information the host or tools actually expose; mark anything else as unverified. Do not ask for passwords or tokens in chat. If you cannot read the guide, send me to https://www.openaikits.com/docs/start.
Roadmap
Where memory is going.
Building on the available OAK.memory engine and cloud connections, we are exploring what comes next. The initiatives below describe future plans.
Roadmap
Cloud Memory for Enterprise
Managed, secure memory infrastructure for teams and their agents, at scale.
Roadmap
Memory-native Assistant
A personal assistant that actually remembers you — across every session and device.
Research
Human Memory Cloning
Capture and reconstruct a person's associative memory graph.
In development
Action Foundation Model
A foundation model with memory-manipulation-optimised embeddings and native actions.
Manifest
Why we build this.
Our Vision
Memory belongs in the stack as infrastructure — the way inference and model providers already are. General-purpose and domain-independent, not re-invented inside every application.
Our Mission
To deliver state-of-the-art AI capabilities with unmatched convenience and cost — so anyone can give their AI a real, persistent memory without the complexity.
About
What we optimize for.
01
Innovation
We turn the latest AI research — associative memory, hybrid retrieval — into practical, drop-in tools.
02
Accessibility
Start with cloud connections or a local engine. Follow guides for installation, store connection, and verification.
03
Efficiency
Compose context with vector search and graph traversal. Choose embeddings and storage for your environment.
Let's talk.
Building with OAK.memory, or want it in your stack? Send a note.
Maintainer
Independently built and maintained.
