AI integration

Our documentation is already prepared for AI agents. Let Claude, Cursor or any Model Context Protocol (MCP) client write the code for you — reading the API reference in real time, without ever leaving the editor.

Why use AI?

Integrating payments means reading a lot of documentation. Authentication, order creation, card tokenization, webhooks, error codes, rate limits, installment math...

With the official Appmax MCP, your agent handles all of that on its own:

• Less back and forth — the agent queries the documentation on its own while writing code
• Instant diagnostics — HTTP errors identified immediately, with root cause and suggested fix
• Ready-to-use code — snippets in curl, Node, Python, PHP, and Go with environment-specific URLs
• Guided onboarding — personalized checklists and typed webhook schemas

In action

Build from scratch in minutes

Ask for a full checkout integration — customer, order, Pix and card payment — and the agent writes everything while querying the Appmax API in real time.

Automatic error diagnosis

The agent hits an HTTP 401 on /v1/customers. Instead of guessing, it calls diagnose_error and pinpoints the cause: app credentials were used where merchant credentials were expected. Fix in seconds.

Webhooks with no surprises

From "I need to receive payment confirmation", the agent queries the typed schema for order_approved via get_webhook_schema and builds a complete handler with correct types.

Precision payload edits

"Add Pix as a payment method in my current flow" → the agent validates the payload with validate_payload before sending, and makes the right diff preserving the rest.

What your agent can do

There are 13 tools the agent calls on its own during the conversation. Here is a summary by category — for the full table with descriptions and detailed usage examples, see the setup page.

Documentation

• search_docs — semantic search across the documentation
• list_pages — list available pages with prefix filter
• get_page — get the full content of a page
• get_full_docs — entire documentation as plain text
• check_health — server status and stats

Diagnostics

• diagnose_error — HTTP error root cause (401/422/429/500)
• validate_payload — validate JSON payload against endpoint schema
• validate_order_total — verify order total calculation
• validate_installation_flow — audits the installation flow implementation from project snippets

Code generation

• generate_code_snippet — snippets in curl/Node/Python/PHP/Go
• get_integration_flow — step-by-step flow (checkout, installation, subscription)

Onboarding and webhooks

• get_onboarding_checklist — admin checklist by integration type
• get_webhook_schema — typed schema + example payload for 28 events

Compatible clients

Any client that implements the Model Context Protocol:

• Claude Code and Claude Desktop
• Cursor
• Windsurf
• VS Code (MCP extension / Copilot)
• Cline, Aider, Continue and others

Open standard, no lock-in

Our server implements the Model Context Protocol — the open standard maintained by Anthropic for connecting AI agents to external data sources and tools. Any compatible client connects without proprietary adapters.

Next steps

Set up the MCP

Full configuration, tool table, and detailed usage examples.

Quickstart

Prefer to write code by hand? Jump straight into the quickstart.