# Contents
The trust problem Stripe as the trust layer What this means for credentials Plan purchase Credential rotation What this unlocks for agents Getting started References

Agents are getting good at writing code. The bottleneck has shifted: it's no longer about generating code, it's about having somewhere to run it safely. And more importantly, about getting there without a human navigating dashboards and copying keys between tabs.

The trust problem

When an agent needs compute, someone has to create an account, generate credentials, and wire them into the right place. Today that "someone" is usually a developer doing it manually. The agent can't do it alone because there's no standard way for it to say "I need a sandbox" and get back real, working credentials from a provider it's never interacted with before.

This is fundamentally a trust problem. The agent needs to provision resources from a third-party provider, and both sides need guarantees: the provider needs to know who's paying, and the agent needs credentials it can actually use. There hasn't been a clean protocol for this. Until now.

Stripe as the trust layer

Stripe Projects solves this by acting as the trust authority between developers, agents, and service providers. Stripe already knows who you are, has your payment method, and can vouch for you to any provider in the network. One CLI command handles account creation, authentication, resource provisioning, and credential delivery.

1stripe projects add daytona/sandbox

No account signup flows. No API key generation in a dashboard. No copying secrets into .env files and hoping they don't leak. The agent or developer runs a command and gets back working credentials in their environment.

What this means for credentials

If you've worked on a team shipping AI-powered features, you've seen the credential mess. Keys in Slack threads. Secrets in shared .env files that nobody rotates. Onboarding a new engineer means a scavenger hunt for "which key goes where." Multiply that by every provider your stack depends on.

Stripe Projects treats credentials as first-class infrastructure. They're delivered securely, scoped correctly, and rotatable from the CLI. When someone joins the team or switches machines, stripe projects env --sync gives them exactly what they need without exposing secrets more than necessary.

Plan purchase

Daytona runs on usage-based compute pricing. Before provisioning a sandbox, you activate a credit plan through the same CLI:

1stripe projects add daytona/top-up-0025

Stripe handles the payment via a Shared Payment Token(SPT) which means your payment method stays with Stripe, Daytona never sees your card details. Plans renew automatically, and you can switch between them anytime:

1stripe projects update daytona/top-up-0025 top-up-0500

Credential rotation

Credentials get stale. People forget to rotate them, or nobody wants to touch a key that "just works." Stripe Projects makes rotation a single command:

1stripe projects rotate daytona/top-up-0025
2stripe projects env --sync

New API key issued, old one revoked, environment updated. No manual steps, no downtime.

What this unlocks for agents

Once a plan or sandbox is provisioned, the credentials are synced to your environment. Each provisioned service gets its own environment variable (e.g. DAYTONA_PLAN_API_KEY, DAYTONA_SANDBOX_API_KEY). An agent can pick up any of these and start working with the full Daytona CLI:

1source .env
2daytona login --api-key=<YOUR_DAYTONA_API_KEY_ENV_VAR>
3daytona sandbox create

Replace <YOUR_DAYTONA_API_KEY_ENV_VAR> with the actual environment variable name from your .env depending on which service you provisioned.

From here the agent has full access: creating and managing sandboxes, executing commands, SSH access, taking snapshots, managing volumes, and controlling sandbox lifecycle. The Stripe Projects provisioning step is the bridge that gets the agent from "I need compute" to "I have a fully authenticated Daytona environment" without any human in the loop.

For CLI installation and full documentation: https://www.daytona.io/docs/en/getting-started/#cli

Getting started

1brew install stripe/stripe-cli/stripe && stripe plugin install projects
2stripe projects init my-app
3stripe projects add daytona/sandbox
4stripe projects env --sync

This is a developer preview and we're looking for feedback, especially from teams building AI agents at scale. Try it and let us know what you think at support@daytona.io.

References