How MNUERON pulls memory into your AI

MNUERON is a memory layer. It sits beside the AI tools you already use (claude.ai, ChatGPT, Claude Desktop, Cowork, Claude Code, Cursor, IDE plugins) rather than inside any of them. You don't replace claude.ai or build a new chat box — MNUERON injects the right memory into the AI you already use, at the moment it's needed.

There are three ways that happens. Pick whichever fits the situation.

The three pull modes

Modes 1 and 2 use the Chrome extension. Mode 3 uses the MNUERON MCP server. You can use any combination — they don't conflict.

---

Mode 1 — Manual recall (the easiest test)

Takes about 3 minutes to verify end-to-end.

Setup

1. Install the Chrome extension. Either from the Chrome Web Store listing, or as an unpacked extension: chrome://extensions/ → enable Developer mode → "Load unpacked" → point at the extension folder.
2. Click the MNUERON icon in your browser toolbar.
3. Switch the toggle to Hosted mode.
4. Sign in with your mnueron.com account.

Save a memory to recall later

Go to https://mnueron.com/dashboard → click + New memory → paste something memorable, e.g.

My favorite competitor-analysis framework is the Wedge: pick a product wedge older competitors can't copy because of architecture or business model lock-in.

Save.

Test the recall loop

1. Open https://claude.ai in a new tab. Start any conversation.
2. Click the MNUERON icon in the toolbar to open the Recall panel.
3. Search for something from your memory (e.g. competitor wedge).
4. Click the result → Inject into prompt.
5. The text appears in claude.ai's prompt box, prefixed with a Context from MNUERON: block.
6. Type your follow-up question and send. Claude answers using the memory you injected.

That's the loop working end-to-end.

---

Mode 2 — Ambient suggest

Passive variant of Mode 1: MNUERON watches what you're typing and surfaces relevant memories above the prompt box without you opening the panel.

Setup

1. Open the MNUERON extension Options page (right-click the icon → Options).
2. Toggle Ambient context on.
3. Optional: pick a namespace scope (default: all).

Test

1. Open https://claude.ai.
2. Start typing a question that relates to a memory you've saved.
3. A floating strip appears above the prompt box with 1–3 matching memories.
4. Click one → it injects.
5. Send.

If nothing appears, your saved memories may not be semantically close to what you're typing. Try a query that quotes phrases from the memory's content.

---

Mode 3 — Agentic (MCP)

The most powerful mode. You don't click anything; the AI itself decides when to query MNUERON. Works in Claude Desktop, Cowork, Claude Code, Cursor, and any other tool that supports the Model Context Protocol.

Prerequisites

• Node.js 18+ on PATH
• A clone or install of MNUERON (npm install -g mnueron, or a local source checkout)
• A fresh API token from https://mnueron.com/account-settings/tokens named something like claude-desktop-mcp

Configure Claude Desktop (Windows example)

Edit %APPDATA%\Claude\claude_desktop_config.json. If the file doesn't exist, create it. Add the mnueron entry to mcpServers:

{
  "mcpServers": {
    "mnueron": {
      "command": "node",
      "args": ["C:\\Mnueron\\Mnueron\\mnueron-v0.1.0\\mnueron\\dist\\index.js"],
      "env": {
        "MNUERON_API_URL": "https://www.mnueron.com",
        "MNUERON_API_TOKEN": "mnu_YOUR_FRESH_TOKEN"
      }
    }
  }
}

Replace the path with wherever MNUERON is built, and the token with the one you just created.

If you installed via npm (npm install -g mnueron), the package exposes a mnueron-mcp binary and you can simplify to:

{
  "mcpServers": {
    "mnueron": {
      "command": "mnueron-mcp",
      "env": {
        "MNUERON_API_URL": "https://www.mnueron.com",
        "MNUERON_API_TOKEN": "mnu_YOUR_FRESH_TOKEN"
      }
    }
  }
}

Mac / Linux paths

The config file lives at:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Linux: ~/.config/Claude/claude_desktop_config.json

The JSON contents are identical.

Restart Claude Desktop

Important: fully quit Claude Desktop. Closing the window isn't enough — right-click the system tray icon and choose Quit (or Cmd+Q on Mac). Then relaunch.

Verify the connection

In Claude Desktop, start a new conversation and ask:

What tools do you have available right now?

You should see a list including these MNUERON MCP tools:

Memory tools

• memory_save
• memory_recall
• memory_get
• memory_get_thread (local mode only — see troubleshooting)
• memory_list
• memory_delete
• memory_namespaces
• memory_import_chat
• memory_import_cowork

Procedural (runbook) tools

• procedural_list — browse saved runbooks
• procedural_match — find a runbook by trigger phrase
• procedural_get — fetch a runbook's full steps
• procedural_record_outcome — log success/failure after running one

Test the agentic loop

Save a clearly identifiable memory in your dashboard first, then ask Claude in Desktop:

What is my favorite competitor analysis framework?

Watch the response — you should see a tool-use indicator (an expandable "Used memory_recall" block) where Claude calls the search, sees the results, and then writes the answer using what it found.

That's Mode 3 working end-to-end. No popup, no copy/paste — the AI just knows.

---

How to know which mode to use when

• Quick one-off: Mode 1 (manual recall). You know what memory you want.
• Browsing / drafting: Mode 2 (ambient suggest). You don't know what's relevant yet but want to see options as you type.
• Long-running agentic work (coding sessions, research deep-dives, ongoing tasks): Mode 3 (MCP). The AI saves and recalls memory at the right moments without you context-switching.

Most users end up running all three: ambient suggest in browser chats, MCP in their IDE and Cowork, and manual recall when they want surgical control.

---

Troubleshooting

---

Privacy note

In Local mode, memories live on your machine in ~/.mnueron/memories.db and never leave. The Chrome extension can still capture from claude.ai and chatgpt.com — but it writes to your local MCP server, not our hosted backend.

In Hosted mode, memories live in MNUERON's Postgres on Vercel and are scoped to your org. Row-level security policies prevent cross-tenant access.

You can switch between modes anytime in the extension options — and migrate your local data to hosted (or back) with mnueron migrate.

---

Last updated 2026-05-25