Getting started

1. Store a test key

@klappay/mcp reads the same ~/.klap/config.json the Klap CLI writes, so the key never has to appear in your MCP client's config file:

npx @klappay/cli login --api-key - --base-url https://api.klap.example

Paste a klap_test_... key on stdin and press Ctrl-D. Reading it from stdin (--api-key -) keeps it out of your shell history; setting KLAP_API_KEY before running the command works too. --base-url is required. The file is created with 0600 permissions. You can store a live key the same way later; both slots live side by side.

2. Add the server

Claude Code:

claude mcp add klap-test -- npx -y @klappay/mcp@1.0.0

Claude Desktop (claude_desktop_config.json) or Cursor (~/.cursor/mcp.json):

{
  "mcpServers": {
    "klap-test": {
      "command": "npx",
      "args": ["-y", "@klappay/mcp@1.0.0"]
    }
  }
}

Restart the client. Its MCP log should show klap-mcp: test → <your API host>. If the server refuses to start, the same log has a one-line reason; see Configuration.

3. Try it

Ask the assistant things like:

• "Which Klap environment are you connected to?" (klap_status)
• "Create a 25 USD test charge payable in USDC on Base, expiring in 15 minutes." (charges_create)
• "Mark that charge as confirmed in the sandbox, then show me its timeline." (sandbox_trigger, charges_timeline)
• "Which of my webhooks failed deliveries recently?" (webhooks_list, webhooks_list_deliveries)

4. Add live (optional)

Add a second, separate server with KLAP_ENV=live. It is read-only:

claude mcp add klap-live -e KLAP_ENV=live -- npx -y @klappay/mcp@1.0.0

Only add KLAP_MCP_ALLOW_LIVE_WRITES=1 if you really want the assistant to create live charges or re-send live webhooks; read Security first.

Where to go next

• Configuration: every variable and refusal
• Tools: what each tool takes and returns
• Security: what is excluded and why

For LLMs and agents

This site publishes llms.txt, a link index of every doc page, and llms-full.txt, the full content of every doc page in one plain-text file. Point an agent or RAG pipeline at either to give it these docs without scraping HTML. Both regenerate on every deploy, so they never drift from the pages here.