-
-
Notifications
You must be signed in to change notification settings - Fork 2
Getting Started
Back to Home
This page covers all installation paths, configuration for Claude Desktop, Docker usage, and debugging with the MCP Inspector.
- Node.js 24 LTS or later
- npm (included with Node.js)
- Claude Desktop (for MCP integration) or any MCP-compatible client
# Download and install Chocolatey:
powershell -c "irm https://community.chocolatey.org/install.ps1|iex"
# Download and install Node.js:
choco install nodejs --version="24.14.1"
# Verify the Node.js version:
node -v # Should print a Node.js 24.x version.
# Verify npm version:
npm -v # Should print an npm 11.x version.
# Run the packaged neurodivergent-memory CLI without a global install
npx neurodivergent-memory@latest init-agent-kit
# You will be asked if you want to install the package, Y/n?# Download and install nvm:
curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.4/install.sh | bash
# in lieu of restarting the shell
. "$HOME/.nvm/nvm.sh"
# Download and install Node.js:
nvm install 24
# Verify the Node.js version:
node -v # Should print a Node.js 24.x version.
# Verify npm version:
npm -v # Should print an npm 11.x version.
# Run the packaged neurodivergent-memory CLI without a global install
npx neurodivergent-memory@latest init-agent-kit
# You will be asked if you want to install the package, Y/n?For local development:
git clone https://github.com/jmeyer1980/neurodivergent-memory
cd neurodivergent-memory
npm install
npm run buildFor development with auto-rebuild:
npm run watchdocker run --rm -i -e NEURODIVERGENT_MEMORY_DIR=/data -v ndm-data:/data twgbellok/neurodivergent-memory:0.2.0docker pull ghcr.io/jmeyer1980/neurodivergent-memory:0.2.0
docker run --rm -i -e NEURODIVERGENT_MEMORY_DIR=/data -v ndm-data:/data ghcr.io/jmeyer1980/neurodivergent-memory:0.2.0Add the server to your Claude Desktop config file:
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%/Claude/claude_desktop_config.json
{
"mcpServers": {
"neurodivergent-memory": {
"command": "/path/to/neurodivergent-memory/build/index.js"
}
}
}{
"mcpServers": {
"neurodivergent-memory": {
"command": "npx",
"args": ["-y", "neurodivergent-memory@0.2.0"],
"env": {
"NEURODIVERGENT_MEMORY_FILE": "C:\\Users\\YOUR_USER\\.neurodivergent-memory\\memories.json"
}
}
}
}{
"mcpServers": {
"neurodivergent-memory": {
"command": "docker",
"args": [
"run",
"--rm",
"-i",
"-e",
"NEURODIVERGENT_MEMORY_DIR=/data",
"-v",
"C:\\Users\\YOUR_USER\\.neurodivergent-memory:/data",
"twgbellok/neurodivergent-memory:0.2.0"
]
}
}
}Memories are persisted using a snapshot plus write-ahead log model:
~/.neurodivergent-memory/memories.json
and:
~/.neurodivergent-memory/memories.json.wal.jsonl
These files are created on first use and restored on server startup. In practice, explicit configuration is better than relying on home-directory defaults, especially when mixing Docker and native npx clients.
Recommended cross-client rule: point every client at the same canonical snapshot path with NEURODIVERGENT_MEMORY_FILE or the same canonical directory with NEURODIVERGENT_MEMORY_DIR.
As of v0.2.0 the container runs as the node user. Older mounts targeting /root/.neurodivergent-memory are no longer discovered automatically. Re-mount the same host volume either at /data with NEURODIVERGENT_MEMORY_DIR=/data, or at /home/node/.neurodivergent-memory.
Since MCP servers communicate over stdio, standard debugging approaches don't apply. Use the MCP Inspector for interactive debugging:
npm run inspectorThe Inspector will print a URL you can open in your browser to:
- Send tool calls interactively
- Inspect responses in real time
- Test search queries, graph traversals, and memory operations
- Verify your Claude Desktop config is working
Once connected through Claude Desktop or the Inspector, try these operations to verify everything is working:
1. store_memory — Create a test memory in logical_analysis that says what you did and why you did it
2. search_memories — Search for the content you just stored
3. memory_stats — Confirm the memory count is 1
4. retrieve_memory — Fetch the memory by ID
For the first store_memory call, prefer canonical tags and include a reusable reason in the content itself. A good smoke-test example is a short note that records the action plus the underlying principle, rather than only a bare task log.
All operations should complete in under 100ms.
For a deterministic live readiness check of project_id support, run:
npm run smoke:project-id- GitHub Actions runs on Node.js 24 LTS for CI and release automation
- npm publishes use OIDC provenance with
npm publish --provenance --access public - Docker images are built with Buildx, published to Docker Hub, and emitted with SBOM and provenance metadata
- GitHub Actions generates artifact attestations for the npm tarball and the pushed container image digest
- Tagged releases upload the npm tarball, checksums, and attestation bundles as release assets
See also: Architecture · Release-Notes · Experiment-Report