Connect Claude

Hook Claude up to your AssetLab account - claude.ai and Claude Desktop via the remote connector, Claude Code via one command.

All Claude surfaces connect through the AssetLab MCP server at https://mcp.assetlab.ca. You need one thing first: an API key.

Step 0 - create an API key

In AssetLab, an Administrator goes to Settings → API Keys → New key:

  1. Name it for its purpose ("Claude - Maria").
  2. Choose scopes - start with read scopes; add :write scopes when you're ready for the assistant to create and update records. See Tools & scopes.
  3. Copy the key (al_live_...) - it's shown once.

claude.ai (web)

  1. In Claude.ai, open Settings → Connectors → Add connector.
  2. Paste the connector URL: https://mcp.assetlab.ca
  3. When the AssetLab authorization page appears, paste your API key.
  4. Done - AssetLab tools appear in your conversations. Try: "List my sites."

Claude Desktop

Same flow as claude.ai - add a custom connector with URL https://mcp.assetlab.ca under Settings → Connectors, and authorize with your API key when prompted.

Claude Code

claude mcp add --transport http assetlab https://mcp.assetlab.ca

Claude Code walks through the same authorization (paste your API key), after which AssetLab tools are available in your terminal sessions - useful for scripting bulk operations conversationally.

Advanced: local stdio server

For clients that only speak stdio, the npm package runs the server locally:

npx @assetlab/mcp-server

It requires two environment variables: ASSETLAB_API_KEY (your key) and ASSETLAB_API_URL (the API base URL shown in Settings → API Keys). The remote connector is simpler - prefer it unless you have a reason.

Verifying the connection

Ask something harmless and specific: "How many assets do I have, by site?" A correct, current answer confirms the pipe. If tools don't appear or auth fails:

Disconnecting

Revoke the API key in Settings → API Keys - this severs access immediately, regardless of what any client has cached. Removing the connector in Claude just tidies the UI; revocation is the security action. See Security model.