Skip to main content
This guide gets you from no setup to asking Claude Desktop “who’s overdue on safety training?” in about five minutes.
Prerequisites
  • A Konstantly tenant on Business or Enterprise plan (the MCP Keys settings tab is hidden on Free)
  • Admin or super-user role on that tenant (the mcp_manage permission)
  • Claude Desktop, Claude Code, or Cursor installed
  • Node.js 18+ on your machine (required by npx)

1. Generate an MCP key

In your Konstantly browser tab, navigate to Settings → MCP Keys. If the tab isn’t visible, your tenant’s license doesn’t include MCP access. Talk to your account manager about upgrading. Click Generate key, give it a recognizable name (e.g. “My laptop”, “Production agent”), and choose scopes:
  • Read — always available. Queries, reports, lookups.
  • Write — Business or Enterprise plan. Assignments, user management, course operations.
Click Generate. The full key is shown once — copy it now. You’ll see something like:
This is the only time the full key appears. If you close the dialog without copying, you’ll need to revoke and regenerate. The key acts as you — treat it like a password.

2. Install in your AI tool

Pick the tool you use:
Open your Claude Desktop config file:
  • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
  • Windows: %APPDATA%\Claude\claude_desktop_config.json
Add a mcpServers entry:
Restart Claude Desktop. The 🔌 menu now shows konstantly with its tool list.
Replace both placeholders with your real values:
  • KONSTANTLY_MCP_KEY — the kmcp_... you copied in step 1
  • KONSTANTLY_URL — your Konstantly instance URL, exactly as it appears in your browser (e.g. https://acme.konstantly.com)

3. Ask your first question

Open your AI tool. Try a few prompts that work with the read-only tools:
“Find a user named Sarah.” (fires find_user)
“How many learners have completed the OSHA refresher course?” (fires find_course + get_compliance_status)
“Show me last week’s leaderboard.” (fires get_leaderboard)
“What’s my own training status?” (fires list_my_assignments)
If you have write access, try:
“Assign the new-hire bundle to user 42, deadline in 14 days.” (fires assign_training — writes an audit log entry visible in Settings → Action Log)

4. Verify the connection (optional)

If something’s not working, switch on debug mode by adding to your env config:
Restart your AI tool. The MCP server now logs to stderr — you can see every request, retry, and tool call. Common issues:

5. What’s next