Quickstart

Get started with MemQ

Install, connect, and verify — all in under a minute.

Memory vs. intelligence fabric

Most memory tools store notes. MemQ turns memory into a living system: it writes, searches, reflects, links ideas, protects namespaces, and routes different memory jobs through the right inference path.

How MemQ routes memory work

Fast path — validation, formatting, lightweight summaries, and health checks.

Recall path — semantic search, recent memory, and namespace-aware retrieval.

Reflection path — distillation, contradiction repair, pattern extraction, and Soul Journal consolidation.

High-trust path — billing, VISA, policy, admin, and deployment-sensitive work that fails closed.

Overflow path — provider fallback for resilience; customers do not need to configure it.

Prompts cannot upgrade themselves. VISA entitlement decides access before inference runs.

Prerequisites

A Multinex account — sign up for free
An MCP-capable editor — Claude, Cursor, VS Code, Claude Desktop, or Antigravity

1. Add the MCP config

Paste this into your editor's MCP configuration file. For detailed per-editor instructions, see the full install page.

MCP Config (OAuth)

{
  "mcpServers": {
    "memq": {
      "type": "http",
      "url": "https://mcp.multinex.ai/mcp/v1",
      "transport": "streamable-http",
      "auth": {
        "type": "oauth2",
        "discoveryUrl": "https://billing.multinex.ai/.well-known/oauth-authorization-server"
      }
    }
  }
}

Config file locations

Claude Code — ~/.claude/mcp.json

Cursor — ~/.cursor/mcp.json

VS Code — .vscode/mcp.json

Claude Desktop — ~/Library/Application Support/Claude/claude_desktop_config.json

Antigravity — Settings → MCP Servers

2. Approve the connection

When you start your first session, your browser opens a secure confirmation page at billing.multinex.ai. Sign in (or create an account) and click Approve. This only happens once — your credentials refresh automatically after that.

3. Verify the connection

After approval, your AI auto-discovers all 22 memory tools. Run this verification sequence to confirm everything works:

First-call verification

# 1. Session handshake — confirms auth + namespace
reunion

# 2. Check your namespace and access tier
namespace_info

# 3. Write a test journal entry
journal_record  (content: "First MemQ connection verified")

# 4. Search for it
journal_search  (query: "first connection")

# 5. Pack context for the current session
mnemosyne_context

If all five calls return successfully, you're connected and your AI has persistent memory.

First useful memories

After the test entry works, add one real memory. This gives your agent immediate value and proves the loop before you depend on it.

Good first entries

Coding agent - "This repo prefers strict TypeScript, Zod at IO boundaries, and proof-led benchmark claims."

Research agent - "When comparing memory systems, separate managed-cloud claims from OSS adapter benchmarks."

Team runbook - "For billing usage issues, check reads, writes, segments, source, and last recorded usage event."

Next steps

→

Tool Reference — Learn what all 22 tools do and when to use them

→

Editor Setup — Install extensions, Claude plugins, and munx-cli

→

Teams & Namespaces — Set up shared team memory