Back to Getting Started
General User track · ~10 minutes

Get Mnueron memory into every AI tool you already use.

No code required. We'll install the CLI, wire it up to Claude Desktop and Claude Code, open the dashboard so you can see what's happening, and finish with a two-minute demo that shows your AI remembering something you taught it.

Step 1

Install the Mnueron CLI

The CLI is a small Node program. It detects every AI tool on your machine and configures each one to talk to Mnueron. Open a terminal and run:

npm install -g mnueron
mnueron setup

You'll see a checklist scroll by — "Claude Desktop ✓ added", "Cursor ✓ added", "Windsurf — not detected", and so on. Anything Mnueron found gets wired automatically.

No Node? Install it from nodejs.org first (the LTS version is fine). Apple Silicon Macs may also need brew install node.
What it does behind the scenes:Mnueron edits each tool's MCP config file (e.g. ~/.cursor/mcp.json) to register itself as a memory provider. You can preview the edits without applying them with mnueron setup --dry-run.
Step 2

Restart Claude Desktop + Claude Code and verify the connection

MCP plugins load at app startup, so quit and reopen both Claude Desktop and Claude Code. In a fresh Claude Desktop chat, ask:

What MCP tools do you have available?

You should see six new tools listed: memory_save, memory_recall, memory_list, memory_delete, memory_namespaces, and memory_import_chat. If you see them, the connection works.

Now try Claude Code. Run claude in a project terminal and ask the same question. Same six tools should appear.
Didn't see them? Quit the app fully (it has to fully exit, not just close the window), then reopen. If still missing, run mnueron setup --only claude-desktop to force a reconfigure, then restart.
Step 2.5

Sanity check — confirm each tool is using the mode you want

Two settings can disagree if you're not careful: the cloud dashboard's storage mode(what your org uses) and each AI tool's MCP config file on your machine (what the tool actually calls). The dashboard setting does NOT reach into your laptop and reconfigure Claude. So after Save, run this check.

Authoritative check (one command)

mnueron stats
# First line tells you the mode:
#   mode: local SQLite                        ← local
#   mode: hosted https://api.mnueron.com      ← cloud

Where each tool's MCP config file lives

The mnueron CLI prints the exact paths when you run mnueron setup. The common Windows paths (Microsoft Store install of Claude Desktop) are below — macOS / Linux fall back to the home-directory equivalents.

Claude Desktop  C:\Users\<you>\AppData\Local\Packages\Claude_pzs8sxrjxfjjc\LocalCache\Roaming\Claude\claude_desktop_config.json
Claude Code     C:\Users\<you>\.claude\settings.json
Cursor          C:\Users\<you>\.cursor\mcp.json
Codex           C:\Users\<you>\.codex\config.toml
Cline (VSCode)  C:\Users\<you>\AppData\Roaming\Code\User\globalStorage\saoudrizwan.claude-dev\settings\cline_mcp_settings.json
Continue        C:\Users\<you>\.continue\config.json

# macOS swaps the path roots:
~/Library/Application Support/Claude/claude_desktop_config.json
~/.claude/settings.json
~/.cursor/mcp.json
~/.codex/config.toml
~/Library/Application Support/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json
~/.continue/config.json

How to read which mode a tool is in

Open the file. Look at the mnueron server entry. The discriminator is the env block:

// HOSTED mode — talks to the cloud
"mnueron": {
  "command": "mnueron",
  "args": ["--mcp-stdio"],
  "env": {
    "MNUERON_API_URL":   "https://api.mnueron.com",
    "MNUERON_API_TOKEN": "mnu_..."
  }
}

// LOCAL mode — talks to ~/.mnueron/memories.db
"mnueron": {
  "command": "mnueron",
  "args": ["--mcp-stdio"]
  // no env, or env without MNUERON_API_URL / MNUERON_API_TOKEN
}

How to flip a tool

# All detected tools → local
mnueron setup

# All detected tools → hosted (cloud)
mnueron setup --hosted https://api.mnueron.com --token mnu_xxx

# A specific tool only
mnueron setup --only claude-desktop
mnueron setup --only codex
mnueron setup --only cursor
After any setup change, fully quit the affected tools (not just close the window) and reopen. MCP servers load at app startup; a still-running Claude Desktop is reading the OLD config until you restart it.
Confirm it stuck.In a fresh chat ask: "What memory tools do you have available?" You should see memory_save, memory_recall, memory_list, memory_delete, memory_namespaces, memory_import_chat.
These two steps are saved as runbooks in your Mnueron memory — ask any AI tool "where is the Claude Desktop MCP config?" or "is Claude using local or hosted mnueron?" and the right runbook will surface automatically.
Step 3

Open the dashboard UI to see your memories

There are two dashboards. Local dashboard runs on your machine and shows the memories saved in your local SQLite. Hosted dashboard runs at app.mnueron.com and shows memories synced to the cloud (you'll set that up in step 4).

mnueron dashboard
# opens http://localhost:3122 in your browser

You'll land on the dashboard overview. Empty right now — you haven't saved anything yet. The sidebar has Memories, Chat, Mnueron Meet, Analytics, Entities, and a Settings link. Anything you save from Claude Desktop will show up here in real time.

Hosted dashboard preview. Even without an account, you can browse the look-and-feel by visiting /dashboard/overview on the marketing site (you'll be prompted to sign in; that triggers step 4).
Step 4

Sign up for cloud sync (optional but recommended)

Local-only is great for a single machine. The moment you want your laptop and desktop to share memories — or share with a teammate — you need a hosted account. The Free tier covers personal sync.

  1. Visit /signup and create an account. You can sign up with Google or email + password.
  2. Once you're in, open /account-settings/tokens and click Create token. Copy themnu_… value — you only see it once.
  3. Back in your terminal, point the CLI at the cloud:
    mnueron setup --hosted https://api.mnueron.com --token mnu_xxx
  4. Restart Claude Desktop. New memories now sync to the cloud AND your other devices.
What about my old local memories? They stay on the local SQLite. To bring them up to the cloud, run mnueron migrate-local-to-hosted (warning: bulk upload may take several minutes depending on count).
Step 5

Choose where your memories live

Once you have a hosted account, you have three storage choices. You pick this in /account-settings/storage — the three cards live there.

Local only

Everything stays in ~/.mnueron/memories.db. Nothing leaves your machine. Best for privacy-sensitive work or offline use. The local CLI manages it; the dashboard will show empty stats for devices not syncing.

Mnueron Cloud (default)

We host your memories on managed Supabase, encrypted at rest with RLS per-organization. Daily backups. Free and Basic plans use this mode. You pay for storage as part of the subscription; there's no S3 bill to manage.

Cloud + Archive (Pro+)

Hot tier (recent memories) stays in Mnueron Postgres. Cold tier (older than your threshold) auto-archives nightly to your own bucket — AWS S3, Cloudflare R2, or Backblaze B2. You hold the credentials, you pay the storage bill, you control retention. Setup walkthroughs: S3 · R2 · B2.

To switch modes: click any card on /account-settings/storage — you'll see a green "Storage mode set to X. Saved." pill confirm the change persisted. The Cloud + Archive card expands an inline form to fill in your bucket credentials, with a Test connection button before you save.
Responsibility split. For BYO storage modes, the bucket and its IAM policies are yours to harden. We document the exact split at /docs/byo-storage.
Step 6

Bring your own LLM key (optional, Pro+)

Mnueron uses an LLM for summaries, decision extraction, and the Project classifier. Free and Basic plans use our shared key with reasonable rate limits. Pro and above can paste their own OpenAI or Anthropic key so the bill goes straight to them and there's no shared-pool ceiling.

  1. Open /account-settings/keys.
  2. Click Edit on the OpenAI or Anthropic card. Paste your sk-… key.
  3. Click Test. A green tick means the key works and the upstream vendor accepts it.
  4. Click Save. The key is encrypted server-side and used only when fulfilling your org's requests.
Audit trail. Every time the server decrypts your key (e.g. to call OpenAI on your behalf), a row is written to an append-only audit log. The actor + route + reason are recorded so a security review can show who triggered the use.
Step 7

Optional — turn on Mnueron Meet for meeting capture

If you also want meeting capture (Granola / Fathom / Zoom / Teams), the Meet-specific quickstart at /docs/meet/getting-started takes you through connecting a meeting source, running the first sync, creating a project, and confirming auto-assignments. Five steps, takes about ten minutes.

Meet is built on top of the same Mnueron memory layer you just installed. Meetings become a special kind of memory; the same recall tools find them, projects organize them.
Step 8

The two-minute demo — watch Claude remember

The fastest way to feel why this matters. Open Claude Desktop (or Claude Code) and run this in one chat:

Save this as a memory for my project:
"Always use Tailwind v4 utility classes, never inline styles.
 Prefer rounded-2xl for cards and rounded-md for buttons."

Claude calls memory_save behind the scenes and confirms. Now close that chat and open a brand-new conversation (the equivalent of a fresh session). Type:

What's my CSS preference for this project?

Claude calls memory_recall, finds the convention you taught it five minutes ago, and reads it back. That's persistent memory across sessions, in two API calls, without you wiring anything up per-provider.

What if you switch providers? Repeat the demo in Cursor (which uses Claude or OpenAI underneath, depending on your settings). Same memories, same recall. Mnueron sits under all of them.
What now? Open the dashboard at mnueron dashboard and watch the count go up as you use Claude across your day. The Analytics tab shows what percentage of your AI sessions are recalling vs. starting cold.