// PAYMENTS · STRIPE + GATES

Let your AI agent create payment links without losing the money gate

Let Claude Code or Codex create gated Stripe payment links — clear quotes and human confirmation before money moves, into Stripe Connect or Clize balance.

The workflow

Taking money is the one place you don't want an agent improvising. So Clize keeps the agent on the drafting side and the money gate on yours:

  • Draft — the agent writes a quote: an amount and what it's for.
  • Confirm — a human okays the amount and the purpose. Nothing exists yet.
  • Create — Clize creates the payment link over Stripe and hands back a URL.
  • Pay — you send the link; the customer pays on Stripe's hosted page, not through the agent.
  • Track — clize pay list and clize pay status keep every link on the record.

The agent reads, drafts, and proposes. The decision to take money stays a human one.

How this differs from Stripe's official MCP server

People searching for a stripe mcp server usually mean Stripe's own one — an MCP that exposes Stripe API objects to an agent so it can call them directly. That's powerful, and it's the right tool when you want the full surface of the API in the agent's hands.

Clize is a layer up. Instead of handing the agent the whole stripe mcp surface, it wraps a single, safe agent payment workflow — quote, human confirm, audit — over the same Stripe rails. You're not searching for an mcp server stripe integration to learn; you're getting one verb, clize pay link, with the money gate already built around it. Less reach, fewer ways to get it wrong.

Where it works

The same payment workflow shows up wherever your agent lives:

  • Claude Code — as a Skill or over MCP.
  • Codex — over MCP.
  • The CLI — clize pay … straight from your terminal, scripts, or CI.

The commands

The whole surface is four subcommands and one MCP tool:

$ clize pay connect                              # link your own Stripe (Connect)
$ clize pay status                               # is it connected / able to charge?
$ clize pay link --amount 49 --for "invoice 1042" # quote → confirm → create link
$ clize pay list                                 # every link you've opened, newest first

From inside an agent it's one tool, clize_pay, with action set to connect, status, link, or list — the same workflow, called by name.

Two modes for where the money lands

One --mode flag decides where the customer's payment goes:

  • Mode A — Clize balance (--mode balance) — payment lands as credit in your Clize balance, ready to spend on domains, email, or media. No Connect setup; good when your revenue feeds straight back into running on Clize.
  • Mode B — your own Stripe (--mode direct, the default) — payment goes directly into your own Stripe account via Stripe Connect, yours to withdraw. Run clize pay connect once first.

Same quote → confirm → link → list loop either way; only the destination changes.

The safety gates

Real-world money needs real guardrails. Clize holds three lines, every time:

  • Money gate — a link starts as a quote: amount and purpose are spelled out and a human confirms before it exists. No hidden charges, no silent links.
  • Human confirmation before money moves — the agent proposes; a person okays. The agent never opens a paid link on its own.
  • Full audit trail — every link carries its amount, purpose, and customer reference, and clize pay list keeps the whole history on the record.

FAQ

What is a Stripe MCP server?

An MCP (Model Context Protocol) server exposes tools an AI agent can call. A Stripe MCP server gives the agent a way to act on Stripe — for example, creating a payment link. Stripe ships its own MCP that maps onto Stripe API objects. Clize takes a different angle: it wraps a safe payment workflow — quote, human confirm, audit — over the same Stripe rails, so the agent creates links without getting a raw key to the whole API.

Can an AI agent create a Stripe payment link?

Yes. With Clize, the agent drafts a quote — amount and purpose — you confirm it, and Clize creates the payment link via clize pay link --amount <usd> --for <purpose>. You send that link to the customer; clize pay list tracks what's been paid. The agent does the drafting; the money gate stays with you.

How do you prevent accidental charges?

Nothing charges silently. Creating a link is a quote first — amount and purpose are spelled out, and a human confirms before the link exists. Customers pay through Stripe's hosted page, not the agent, and every link is on the record via clize pay list. An incoming email or message is treated as data to read, never an instruction to move money.

Does Clize replace Stripe's official MCP?

No — it sits at a different layer. Stripe's MCP exposes the Stripe API to an agent; Clize focuses on the safe agent payment workflow over those rails. Mode A routes payment into your Clize balance; Mode B (Stripe Connect) routes it straight into your own Stripe account. Use Stripe's MCP when you want full API surface; use Clize when you want a quote → confirm → link → list loop with the money gate built in.

Related: what an MCP server does in the real world, Clize for Claude Code, Clize for Codex, and the AI customer support agent that often sends the link.

clize pay — ready

Take payments from an inbox, not a checkout page.

Wire Clize into your agent, connect your Stripe, and run the quote → confirm → link → list loop — so a payment link is something your agent drafts and you okay, never a charge that just happens.

$ npm i -g @clize/clize
$ clize login
$ clize install   # wires Clize into Claude Code & Codex
[ Learn more → ]