mnueron's setup wizard auto-detects every AI tool you have installed and
wires the local MCP server into each of them in one shot. You install the
CLI once, run mnueron setup, and any AI tool that speaks the Model
Context Protocol gains the same six memory tools.
What gets wired
| Tool | Detection | What's added |
|---|---|---|
| Claude Desktop | macOS ~/Library/Application Support/Claude/, Windows %APPDATA%\Claude\, Linux ~/.config/Claude/ | claude_desktop_config.json updated with the mnueron MCP server stanza |
| Claude Code | claude binary on PATH | ~/.config/claude-code/mcp.json updated |
| Cursor | ~/.cursor/ exists | mcp.json in the Cursor config dir |
| Windsurf | ~/.codeium/windsurf/ exists | mcp_config.json updated |
| Cline (VS Code) | cline_mcp_settings.json present | The settings file gets the mnueron stanza |
| Continue.dev, Zed, Aider, Goose, OpenCode | Detector planned in v0.2.7 | Same pattern — JSON config update |
Plus two non-MCP surfaces:
| Surface | How it talks to mnueron |
|---|---|
| Chrome / Firefox extension | Talks to the local CLI's HTTP server at http://127.0.0.1:3122 (or hosted URL with a mnu_... token). Captures from claude.ai, chatgpt.com, gemini.google.com. |
VS Code extension (mnueron-vscode) | Native panel — works alongside Copilot / Cline. Talks to either the local CLI or hosted backend. |
One-command install
# Node 20+ required
npm install -g mnueron
# Walk through detected tools, confirm each
mnueron setup
The wizard prints what it found and asks before writing anything. After
each tool restarts, it'll see memory_save, memory_recall,
memory_list, memory_get, memory_get_thread, memory_delete,
memory_namespaces, plus the procedural-memory tools procedural_match,
procedural_list, procedural_get, and procedural_record_outcome.
Verify it worked
Open any wired AI tool and ask:
"What MCP tools do you have available?"
You should see the mnueron tools in the list. If a tool doesn't list them, restart it — most pick up new MCP servers only on launch.
Hosted vs local
By default the wizard configures local mode — everything lives in
~/.mnueron/memories.db on your machine, no account needed.
To wire a tool to the hosted backend instead, pass --hosted and your
token:
mnueron setup --hosted https://www.mnueron.com --token mnu_yourtokenhere
Get a token at https://www.mnueron.com/account-settings/tokens.
Manual config
If your tool isn't auto-detected, paste this into its MCP config:
{
"mcpServers": {
"mnueron": {
"command": "npx",
"args": ["mnueron", "mcp"]
}
}
}
That spawns the mnueron MCP server on demand and points the tool at local SQLite. For hosted, swap to:
{
"mcpServers": {
"mnueron": {
"command": "npx",
"args": ["mnueron", "mcp"],
"env": {
"MNUERON_API_URL": "https://www.mnueron.com",
"MNUERON_API_TOKEN": "mnu_yourtokenhere"
}
}
}
}
What to read next
- SDK overview — wire mnueron into your own app, not just IDEs.
- Chatbot with persistent memory — a working bot pattern using the SDK.
- Complete reference — every capability in one page.