Panaxon AIAI Memory Platform

Developers

M.N.E.M.O. is more than Dorothy.

Dorothy is one interface built on M.N.E.M.O. Underneath, M.N.E.M.O. — short for Multidimensional Neural Elastic Memory Organization — is a vector-free memory and knowledge-sharing layer — you can talk to it directly, connect it to Claude or ChatGPT, call it from inside a Claude Code session, or query it as a RAG pipeline from your own app. Same memory, four ways in.

Build your own interface

The same memory and knowledge-sharing layer that powers Dorothy, available for anything you build.

Ingest documents, spreadsheets, or plain text; ask questions in natural language; get back exact stored values, not paraphrased guesses. Standard, Pro, and Enterprise accounts can create an API token and start integrating — no separate approval step required.

  • Store and recall memory across sessions, not just one conversation window
  • Ask questions in plain language — M.N.E.M.O. figures out what you mean and returns the exact stored fact
  • Organize memory automatically into subjects, so related information stays together without manual folder management
  • Role-based access built in, so sensitive data can be scoped to the right people as your team grows

Available on Standard, Pro, and Enterprise

See plans

Give Claude or ChatGPT a memory

Add M.N.E.M.O. as a connector. Nothing to install, no API key.

M.N.E.M.O. runs as a hosted MCP server, so any MCP-capable chat app can use your memory. Once connected, you can ask the assistant to remember something in a conversation and recall it weeks later in a completely different session. The assistant decides when to save or retrieve; you don't have to invoke anything manually.

  • 1. Claude: in Claude (web, Claude Desktop or mobile) open Settings → Connectors → Add custom connector. Enter https://mnemo-mcp.panaxon.ai/api/mcp and leave any client ID, client secret and header fields empty.
  • 2. ChatGPT: go to chatgpt.com/plugins, select the plus button and choose Add custom MCP server. Give it a name and description, enter the same URL as the connection, choose OAuth when it asks how to authenticate, accept the risk warning and select Create as a plugin. Then install it and start a new conversation. Whether you can add custom MCP servers depends on your ChatGPT plan and workspace policy; see OpenAI's instructions.
  • 3. Sign in: a browser window opens. Sign in to M.N.E.M.O., check the username shown, and click Allow.
  • 4. Try it: "List my M.N.E.M.O. subjects" or "Remember that our Q1 target is $5M".
  • Four tools: ask a question of your memory, remember text or a decision, list your subjects, and check a background store
  • The connector acts as the user who clicked Allow, with that user's role, and only ever reaches that user's own account
  • Connected apps are listed under Security → Connected Apps in Dorothy, and each can be revoked instantly
  • Prefer to run it yourself? An optional local MCP server uses your API key instead

Available on Standard, Pro, and Enterprise

See plans

Bring memory into your coding workflow

Nothing extra to install or keep running — Claude Code talks to your M.N.E.M.O. account directly.

For Claude Code, M.N.E.M.O. is available as a Skill instead of a running server — Claude Code reads a set of instructions and calls it directly. Useful for carrying project decisions, specs, or context across coding sessions instead of re-explaining them every time.

  • No server process to install or manage — Claude Code uses its own built-in tools
  • Ask Claude Code to remember a decision, spec, or piece of context from your current session
  • Pull that memory back up in a future session, on a different day, without re-pasting anything

Available on Standard, Pro, and Enterprise

See plans

A third retrieval paradigm

MnemoRAG: VectorRAG, GraphRAG, MnemoRAG — plug M.N.E.M.O. into your own RAG pipeline.

Most RAG pipelines retrieve from a vector database or walk a knowledge graph. MnemoRAG is a third option: one stateless POST call to a named, fixed subset of your own M.N.E.M.O. subjects. The eNN reasons over your data, OAT retrieves it exactly, and one synthesized answer comes back — no chat session, no conversation history, no disambiguation step to manage.

Customer-ownedMNEMO coreOAT retrievalLLM providerPanaxon-hosted storage
How MnemoRAG fits your pipeline: your RAG pipeline calls MNEMO, which ingests and reasons over your data with an LLM provider of your choice, retrieves the exact facts through OAT, and keeps the data in a Panaxon-hosted or customer-owned database.
  • Define a MnemoRAG endpoint from Dorothy Settings — name it, describe it, and choose exactly which subjects it can see
  • Call it with a single request — no session or chat state for your app to track
  • Evidence from every subject in scope is combined into one answer, never silently blended across mutually exclusive numbers
  • Pass optional context of your own to ground or validate the answer, without it ever becoming a retrieval target itself
  • Turn on a full-memory fallback or a general-knowledge fallback per endpoint — both off by default

Available on Pro and Enterprise

See plans

Not just chat

Same memory, four ways in.

Whichever way you connect — API, MCP, Skill, or MnemoRAG — it's the same underlying memory: vector-free, exact recall, built on OAT. Dorothy shows what that feels like in a chat app. These four are how you build your own.

Get API accessSee plans