Challenge

When former Stripe engineer Stanislas Polu cofounded Dust in 2023, he knew from the outset he wanted to run the company's payments, billing and invoicing on Stripe – trusting its technology to scale as the company grew globally. The two companies crossed paths again when Dust set out to let its customers streamline financial operations tasks ranging from invoice generation to customer refunds to subscription adjustments, using specialised AI agents.

To do so, Dust sought out ways to enable agents built on its platform to interact directly with Stripe data securely, at scale and without building custom integrations that would take engineering resources and create technical debt down the road.

"Dust is really meant to be a company's agent operating system," said Polu. "Our mission is to change the way people work. To do that, we need to be at the forefront of building products that make people feel at ease with using agents to do better, faster and more engaging work. With financial operations being a core part of a company's fabric, integrating with Stripe's MCP looked like a perfect way to let fast-moving teams scale themselves, with Dust agents able to tap payment data."

Solution

Dust agents equipped with internal knowledge and the Stripe MCP

Dust agents have access to internal context (e.g. CRM for customer data, knowledge bases for procedures) to take action and to complete tasks. By integrating with the Stripe Model Context Protocol server – a set of tools that AI agents can use to interact with the Stripe API – Dust customers are able to build agents that securely and instantly use internal knowledge and Stripe data to complete high-value financial operations.

To connect to Stripe and make it available as a tool inside Dust, a developer only needs to specify the MCP server and an API key, then configure permissions. Stripe only returns the specific data an agent is authorised to use for its task, enabling Dust to confidently automate sensitive payment-related workflows without creating a risk for the company or its customers. Since security is paramount – especially when it comes to financial data – Dust's platform allows users to remain in the loop by reviewing and validating actions suggested by agents.

Internal usage as the test bed ahead of public launches

As a fast-moving team constantly looking for ways to scale, Dust has built a habit of frontloading internal experimentation ahead of public launches – a practice the company calls "Dust for Dust." As such, the team started by tackling internal bottlenecks in financial operations before making Stripe's MCP server available to its customer base.

Using the Stripe-powered tooling, Dust created an intelligent refund agent that can detect a refund request, review a customer's history, validate the charge in Stripe and use the company's refund policy to determine whether the request should be approved. Once approved, the agent triggers the refund in Stripe and generates a personalised confirmation email.

In addition, Dust built an invoice generator agent that creates invoices and then sends personalised emails explaining the charges. It also created a Stripe agent that can handle a broader range of requests by combining data from Stripe with CRM, email, data warehouses and other sources to help Dust's customer support teams address inquiries more efficiently.

Results

Dust connects its agents to Stripe in minutes

Since Dust's platform was designed to easily create and deploy AI agents, it took less than five minutes to integrate Stripe through its MCP server and give AI agents secure access to payment data.

"Dust already had the functionality to integrate MCP tools into specialised agents, so all we had to do was specify Stripe as the remote server. From there, you can start creating any use-case-specific agent you want," said Frank Aloia, AI engineer at Dust.

Dust support teams save hours with refund automation

With the Stripe-powered refund agent, Dust automated a multistep refund workflow that had included reviewing customer history, validating invoices, issuing refunds and drafting emails. For each refund request, this process could take an hour. Through automation, the refund workflow can now be executed in seconds.

"Instead of drowning under a long-tail of low value tasks, teams get to maximise time spent on critical work and customers can feel it," said Nicolas Chinot, US general manager at Dust.

Dust is enabling its customers to build AI agents powered by Stripe's MCP

On the heels of success with internal use cases, Dust is preparing to enable thousands of its customers to build agents powered by Stripe's MCP. They will be able to connect their own Stripe account inside Dust, configure permissions and start automating workflows such as refunds, invoicing, account updates and more.

"Companies – ourselves included – are actively looking for ways to ensure that internal bottlenecks do not grow in lockstep with their customer base. Agents powered by internal information, including financial data via Stripe's MCP, act as a powerful lever to ensure maximum scalability," said Chinot.