AI developer tools
Integrate faster with your AI assistant
A public, read-only MCP server brings Localpayment’s documentation into your IDE and AI tools, so generated code starts from the real API contract instead of guesses.
- Public and read-only
- No authentication
- Works inside your IDE
Public and read-onlyToolkit
Documentation built for people and machines
MCP serverPublic
Connect https://docs.localpayment.com/mcp to Cursor, VS Code with Copilot, Claude, Windsurf, ChatGPT, Gemini CLI or JetBrains IDEs.
MCP server —llms.txt
A machine-readable index of the documentation that agents can load in a single request.
llms.txt —Markdown docs
Add .md to any docs URL to get a clean Markdown version, ideal as context for a prompt.
Markdown docs —Prompt library
Ready-made prompts for common integration tasks, written against the real API.
Prompt library —Integrator and QA agents
Agent templates for the OpenAI Agents SDK, LangChain, Vercel AI SDK, CrewAI and Amazon Bedrock.
Integrator and QA agents —Ask AI
Ask questions in plain language right in the documentation and get answers grounded in the docs.
Ask AI —
Setup
Add the MCP server in one step
Point your client at the remote endpoint. There is no API key to manage, because the server only reads public documentation.
- Endpoint: https://docs.localpayment.com/mcp
- Read-only: it cannot see your account, credentials or transactions
- Instructions for every supported client are in the MCP setup guide
{
"mcpServers": {
"localpayment-docs": {
"url": "https://docs.localpayment.com/mcp"
}
}
}Workflow
From prompt to tested integration
- 01
Connect
Add the MCP server to your IDE or AI assistant.
- 02
Ask with context
Describe the flow, country and language you need, for example a Pix pay-in on Stage in Node.js.
- 03
Review
Check the generated code against the API reference and have a teammate review it before merging.
- 04
Test on Stage
Run it with fictitious data, verify webhooks and cover failure paths before go-live.
Prompts to try
Start with a good question
Paste these into an assistant connected to the MCP server, then adapt them to your stack.
Create a pay-in
“Using the Localpayment docs, write a Node.js function that creates a Pix pay-in on Stage with a UUID externalId and handles 409 Conflict.”
Verify webhooks
“Write an Express endpoint that verifies the Localpayment x-Signature header with a timing-safe HMAC-SHA256 comparison.”
Test virtual accounts
“Show me how to create a virtual account in Mexico and simulate an incoming transfer on Stage.”
Handle errors
“Map Localpayment status code families to wait, fix, retry or reverse actions in TypeScript.”
For agents and RAG
Plain text your tools can read
Feed the documentation to your own agents, retrieval pipelines or internal assistants without scraping HTML.
- llms.txt is an index of the documentation designed for language models
- Every docs page is also available as Markdown
- Both are public and need no authentication
# Machine-readable index of the documentation
curl https://docs.localpayment.com/llms.txt
# Any docs page as clean Markdown: add .md to its URL
curl https://docs.localpayment.com/docs/webhooks.mdUse AI safely
Guardrails for AI-assisted integration
AI speeds you up. These habits keep your integration, and your customers, safe.
Never paste secrets
Keep API credentials, tokens and webhook secrets out of prompts, chats and screenshots.
No real customer data
Use fictitious names, documents and card numbers in prompts and tests.
Assume it can be wrong
Assistants can invent fields or endpoints. Check every request against the API reference.
Human review before merge
Treat AI output like any other code: reviewed and approved by someone other than the author.
Read-only by design
The MCP server only reads public docs. It cannot move money or access your account.
Stage first
Run generated code against Stage with test data before it ever touches production.
AI tools FAQ
Does the MCP server need an API key?
No. It is public and read-only, and serves only the public documentation. It never has access to your credentials, account or transactions.
Which tools are supported?
Clients that support remote MCP servers, including Cursor, VS Code with Copilot, Claude, Windsurf, ChatGPT, Gemini CLI and JetBrains IDEs. Product names are trademarks of their respective owners and are listed for compatibility only.
Can an AI agent make payments through it?
No. The MCP server exposes documentation only. Payments always go through the authenticated API, called from your own servers.
Is AI-generated code production-ready?
Not by default. Review it, test it on Stage and follow the go-live checklist, exactly as you would with code written by hand.
What is llms.txt?
A plain-text index of our documentation designed for language models. Your agents can fetch it, then load individual pages as Markdown by adding .md to their URL.
Should we use the MCP server or llms.txt for an internal assistant?
For people coding in an IDE, the MCP server. For your own agents, RAG pipelines or support bots, fetch llms.txt and the Markdown pages. Many teams use both.
Two ways in
MCP server in your IDE, or llms.txt for your own agents
Both are public, read-only and need no authentication. Pick by where the questions are asked.
MCP server: inside your IDE
- Live, question by questionYour assistant fetches the exact page it needs while you code.
- Works in supported clientsCursor, VS Code with Copilot, Claude, Windsurf, ChatGPT, Gemini CLI and JetBrains IDEs.
- Best for day-to-day buildingGenerating requests, webhook handlers and error handling against the real contract.
llms.txt and Markdown: for your own agents
- The whole index in one requestFetch llms.txt, then load any page as Markdown by adding .md to its URL.
- Fits RAG and internal assistantsFeed retrieval pipelines, support bots or integrator agents without scraping HTML.
- Best for automationQA agents, documentation checks and tooling that runs outside the IDE.
Build with AI, ship with confidence
Connect the MCP server, generate your first request and test it on Stage.