Stripe Projects Expands Agent Integrations: AI Agents Provision Infrastructure Safely

Last week, agent traffic surpassed human traffic on the internet for the first time (per Cloudflare Radar). Agents browsing webpages account for part of that, but Stripe data points to a more significant reason: agents building software. Agent traffic to Stripe's documentation grew more than 10x in 2025 and now makes up nearly 40% of docs traffic. Stripe CLI has seen exponential growth in new users, with 70% of CLI requests for API resources coming from agents.

Stripe's data shows agents are now fully capable of independently writing code and integrating with APIs like Stripe's. Yet many of the steps adjacent to writing code — provisioning infrastructure, managing credentials, setting up accounts, and connecting services together — are still too hard for agents to do on their own. Stripe is expanding Stripe Projects, which lets you provision and manage services from the CLI, to solve this.

More agents can provision infrastructure directly

Stripe Projects is now available as a skill in Hermes, the open-source, self-improving AI agent from Nous Research. Projects handles infrastructure provisioning and management, while Hermes carries context across sessions — so if you are building a complex product over days or weeks, your agent is a persistent collaborator, returning to a project the way a developer would. Model-agnostic coding agents Factory Droids and Warp have also integrated with Projects.

An agent is not a throwaway "it runs, done" tool — cross-session context lets it remember architecture decisions, follow existing conventions, and continue where it left off. That also means the credentials and environments it touches accumulate over time. Without environment isolation, an out-of-scope agent could quietly modify production config. This is exactly why Projects makes scoped credentials, spend thresholds, and named environments default capabilities.

16 new providers, 49 in total

Projects now connects to 16 new providers for a total of 49, including Metronome for usage-based billing, Wix for public-facing storefronts, and ClickHouse for LLM observability. Agents can spin up a live AI-powered product, bill users, and monitor model cost, latency, and quality from the first call — without a human clicking through a dashboard.

New developer controls for safe builds

Agents provisioning infrastructure need many of the same guardrails as agents making purchases: scoped credentials, spend thresholds, and enforced environment boundaries. Stripe is shipping three updates:

  • See multiple providers' spend in one place, with current and historical cost per project.
  • Set spend limits per provider — tighter on an AI model provider, higher on production hosting — to prevent an agent from burning the whole budget on one service.
  • Keep builds contained with named environments and isolated credentials for development, staging, or production. Agents default to the development environment, so even an off-script agent won't touch production.

Platforms can also mint scoped credentials and provision infrastructure on their developers' behalf, with white-labeling support, so developers can go from writing code to running it without leaving your environment.

How to use it

Projects is driven from the stripe CLI. The key subcommands look like this:

stripe projects list                 # list projects and provider status
stripe projects connect --provider   # connect a new provider, follow the auth prompt
stripe projects spend                # current and historical spend per provider
stripe projects limits set --provider <name> --monthly <amount>
stripe projects env create staging   # create a named environment with isolated credentials

To budget an agent: run it in the development environment for a week, see what it actually spends, then set the production monthly limit at 1.5x that baseline to leave room for traffic spikes. If a model provider's calls spike abnormally, the spend command pinpoints which agent and which environment are burning money — no need to wait for the month-end invoice.

What's next

Stripe plans to expand Projects across the full lifecycle of agent-built software, from first provision to ongoing operations and security — adding deeper security primitives for autonomous agents and a data layer so providers can meter and bill for the software agents built on their services.

16IDC perspective

For teams using Stripe for SaaS, subscriptions, or platforms, Projects is a reminder that when agents become "developers," cost and permission governance must come first. Get comfortable with Stripe payment integration and subscription billing API integration, and apply the permission model from API security with OAuth/JWT to the credentials you hand your agents. More in the Backend Integration category.

FAQ

Is Projects only for Stripe users? It starts from the Stripe CLI, yes, but the providers it connects to now cover hosting, databases, observability, and more — a good fit for teams building products inside the Stripe ecosystem. What can an agent do with its credentials? Only what the scope and environment allow; by default agents land in the development environment and out-of-scope actions are isolated. Are spend limits hard caps? Yes — requests are blocked once the limit is reached, so the budget cannot be blown through. Does it lock you in? Scoped credentials can be revoked and re-minted at any time, so migration cost stays manageable.

References: Stripe Blog · Stripe Projects adds new agents, providers & developer controls https://stripe.com/blog/stripe-projects-adds-new-agents-providers-developer-controls; Stripe CLI reference https://docs.stripe.com/cli