Run & Compare Agents in one command


Stripe Projects allows you to provision a full Runloop account with devboxes ready for agents, full access to pro plan features, and real credentials in your environment. No dashboard and no copy-paste keys.
We're excited to announce that we are participating in the Stripe Projects developer preview as a design and launch partner. Read on to learn more as to how the new Stripe Projects and Runloop Integration can help you develop durable, agent-driven workflows and multi-agent systems.
If you'd like to skip our writeup and get straight to code, install the Stripe CLI, then run `stripe projects init my-app && stripe projects add runloop`.

Getting AI agents into real, compute environments is still largely a manual process and far from a trivial one. You need to provision and configure infrastructure, generate credentials, figure out where to store them, and finally wire them into your agent's environment. You then do it all again when you need a clean slate.Â
This is a process best delegated to managed providers like Stripe and Runloop.
The entire end-to-end workflow is brittle for humans and is essentially broken for agents. An agent needs deterministic steps and real credentials it can actually use, not just a sequence of dashboard clicks and copy-paste keys.
The same problem shows up when teams want to run benchmarks or evals at scale. Standing up the right compute, getting credentials into the right place and making it repeatable across runs: it's all setup work that gets in the way of the actual work that needs to be done to move the needle on ROI metrics.

Stripe Projects is a new workflow in the Stripe CLI that provisions real services in a user's own provider account and returns credentials in an agent-ready format, all from a single command.
We co-designed the Agentic Provisioning Protocol with Stripe specifically so that this works reliably across many providers. The provisioning steps are deterministic, the credential handoff is secure, and the resources live in the user's account with normal dashboard access preserved.
When you run `stripe projects init my-app` and select Runloop, you get a fully configured Runloop account with all features enabled:
The whole account lives in your name with full dashboard access. Runloop doesn't disappear behind an abstraction layer: you get a fully integrated account with real resources you control.
The current default for getting credentials into an agent workflow is still mostly copy-paste and .env files. It’s not good enough when agents are executing real-world tasks and deployed at scale. Stripe Projects gives us a clean, auditable path for credential delivery that we can trust.
Unlike generic code execution sandboxes or containers, Runloop devboxes are natively built for agents. It made sense to partner on a provisioning protocol that matches: deterministic, scriptable, and reliable enough that an agent can drive the whole flow end-to-end.
Repeatable setup unlocks real eval pipelines. Teams running benchmarks need to stand up clean environments repeatedly and predictably. Combining Runloop's cloud orchestrator with Stripe Projects' provisioning workflow makes that dramatically simpler.
Resources provisioned through Stripe Projects live in the user's own Runloop account with standard access. There's no lock-in and no black box.
“Stripe Projects is a natural fit for Runloop. Our platform provides AI agents with secure cloud environments to build, test, and deploy code. Stripe Projects lets developers onboard onto Runloop and manage their account with the same payments and billing infrastructure they already trust. Together, we're making it dramatically easier to ship production-grade AI applications.”
- Jonathan Wall, Founder & CEO, Runloop.ai Â

In order to quickly get yourself up and running with Stripe Projects and Runloop, follow the steps below. If you'd like additional guidance and prefer video, check out our quick tutorial linked below.
Interactive Tutorial: https://youtu.be/89qkfRFWhy0
Docs: https://docs.runloop.ai
This is a developer preview and we're actively looking for feedback, especially from teams building agent workflows or running evals at scale. Send your feedback to dev@runloop.ai!