Skip to content

Quick Start

This guide walks you through basic NeuralMemory usage in 5 minutes.

3 tools you need

NeuralMemory has 63 tools, but you only need three: nmem_remember, nmem_recall, and nmem_health. The agent handles the other 60 automatically. See all tools.

0. Setup

Claude Code (Plugin)

/plugin marketplace add nhadaututtheky/neural-memory
/plugin install neural-memory@neural-memory-marketplace

OpenClaw (Plugin)

pip install neural-memory
npm install -g neuralmemory

Then in ~/.openclaw/openclaw.json:

{
  "plugins": {
    "slots": {
      "memory": "neuralmemory"
    }
  }
}

Restart the gateway. The plugin auto-registers 6 tools (nmem_remember, nmem_recall, nmem_context, nmem_todo, nmem_stats, nmem_health) and injects memory context before each agent run. See the full setup guide.

Cursor / Windsurf / Other MCP Clients

pip install neural-memory

Then add nmem-mcp to your editor's MCP config. No nmem init needed — the MCP server auto-initializes on first use.

VS Code Extension

Install from the VS Code Marketplace for a visual interface — sidebar memory tree, interactive graph explorer, CodeLens on functions, and keyboard shortcuts for encode/recall.

Optional: Explicit Init

nmem init    # Only needed if you want to pre-create config and brain

Your First 5 Minutes (MCP)

If you're using Neural Memory through an AI agent (Claude Code, Cursor, etc.), here's the realistic flow:

Minute 1 — Install and restart your AI tool. The MCP server auto-initializes.

Minute 2 — Store your first memories by telling the agent:

"Remember: we chose PostgreSQL over SQLite because we need concurrent writes"

"Remember: the auth bug was caused by a null token in the session middleware"

Minute 3 — Recall naturally:

"What did we decide about the database?"

"What was that auth bug about?"

Minute 4 — Check brain health:

"Run nmem_health and show me the report"

Minute 5 — That's it. The agent handles sessions, context loading, and maintenance in the background. Keep building — memories accumulate and get smarter over time.

CLI Usage

The following sections show CLI commands for direct usage. If you're using MCP (above), the agent calls these automatically.

1. Store Your First Memory

nmem remember "Fixed auth bug with null check in login.py:42"

Output: 

Stored memory with 4 neurons and 3 synapses

2. Query Memories

nmem recall "auth bug"

Output: 

Fixed auth bug with null check in login.py:42
(confidence: 0.85, neurons activated: 4)

3. Use Memory Types

Different types help organize and retrieve memories:

# Decisions (never expire)
nmem remember "We decided to use PostgreSQL" --type decision

# TODOs (expire in 30 days)
nmem todo "Review PR #123" --priority 7

# Facts
nmem remember "API endpoint is /v2/users" --type fact

# Errors with solutions
nmem remember "ERROR: null pointer in auth. SOLUTION: add null check" --type error

4. Get Context

Retrieve recent memories for AI context injection:

nmem context --limit 5

With JSON output for programmatic use:

nmem context --limit 5 --json

5. View Statistics

nmem stats

Output: 

Brain: default
Neurons: 12
Synapses: 18
Fibers: 4

Memory Types:
  fact: 2
  decision: 1
  todo: 1

6. Manage Brains

Create separate brains for different projects:

# List brains
nmem brain list

# Create new brain
nmem brain create work

# Switch to brain
nmem brain use work

# Export brain
nmem brain export -o backup.json

7. Web Visualization

Start the server to visualize your brain:

pip install neural-memory[server]
nmem serve

Open http://localhost:8000/ui to see:

  • Interactive neural graph
  • Color-coded neuron types
  • Click nodes for details

Example Workflow

Here's a typical workflow during a coding session:

# Start of session - get context
nmem context --limit 10

# During work - remember important things
nmem remember "UserService now uses async/await"
nmem remember "DECISION: Use JWT for auth. REASON: Stateless" --type decision
nmem todo "Add rate limiting to API" --priority 8

# When you need to recall
nmem recall "auth decision"
nmem recall "UserService changes"

# End of session - check what's pending
nmem list --type todo

Next Steps