Meet PontisGlobe AI Agent - Aurora
Integrate in minutes.
Without writing code.
Connect your AI assistant to PontisGlobe MCP and describe what you want in natural language. Aurora reads our reference, writes the code, and checks every request against the same validators our API runs — before you send anything.
Free · Connects in under a minute
Ask for what you want. Watch it check its own work.
Aurora does not guess at our API. It calls the tools, reads the real answer, and fixes its own mistakes before a single request leaves your machine.
Add PontisGlobe to my Node app — collect USDT from my customers, and pay out 250 USDT to a Nigerian bank account.
get_coverageAfrica
Nigeria · NGN · 7 countries in region
get_recipient_fieldsbank_local
required: account_name, account_number · 3 optional
validate_payout_requestdraft
1 problem — idempotency_key is required
validate_payout_requestdraft, corrected
valid — matches the live payout schema
validate_payin_requestcheckout draft
valid — matches the live payin schema
Both directions wired — signed, typed and validated before a single live call.
What your assistant can do
Eight tools, ready the moment you connect. Four answer questions about the product; four check what you are building.
Answer real questions
Pricing, settlement speed, coverage, refunds, onboarding, compliance and contracting — answered from what we publish, not from what a model half-remembers about payments.
Build the request
Recipient fields differ per payment method, and getting them wrong is the most common reason a first payout is rejected. Aurora looks them up instead of guessing.
Check it against the real thing
The validators behind these tools are the exact ones our API runs. A request that passes here will not be rejected for its shape.
Write the integration
Signing, the request, callback verification — in JavaScript, Python, Go, Java, Ruby, PHP or C#. Ask it to port an integration between them and it will.
Three steps, about a minute
- 01
Add Aurora
One command in Claude Code or Codex, one line of JSON in Cursor, VS Code or Windsurf.
- 02
Ask in natural language
No syntax to learn and no slash command. Your assistant works out which tool to call on its own.
- 03
Ship it
Test everything against the sandbox with no real money, then point the same code at production.
Works with the tools you already use
Anything that speaks MCP over HTTP. Setup for each is in the guide.
- Claude Code
- Claude Desktop
- Codex CLI
- Cursor
- VS Code
- Windsurf
- Gemini CLI
- Cline
- ChatGPT
Safe to connect
Public on purpose. Empty by design.
There is nothing behind these tools that belongs to you. They cannot read a balance, look up a payment, list your recipients or move money — they answer questions about the product and the shape of a request. Account-aware tools are coming, and those will ask you to connect your account first.
- Aurora reads none of your data — it has no way in
- Nothing you send is stored. There is no database behind it
- Your prompts and drafts never reach a log line
- Read-only — it cannot move money or read a balance
- Rate limited, size bounded, and origin checked
- What we do count: which client connected, and which tool it called. Never the content of either.
Point your assistant to our MCP
It takes one command, and you can ask Aurora anything before you decide whether to build.