The Rise of the Agentic Developer

The internet is shifting from human-driven browsing to agent-driven execution. Stripe data shows that agent traffic to its documentation grew more than 10x in 2025, now accounting for nearly 40% of documentation traffic. This is not merely about automated reading; it is about automated building.

The shift is visible in the command line. Currently, 70% of Stripe CLI requests for API resources originate from agents. While agents can now independently write code and integrate with APIs, they still struggle with the operational overhead of software development—provisioning infrastructure, managing credentials, and connecting services.

To bridge this gap, Stripe is expanding Stripe Projects to provide the necessary infrastructure management for autonomous workflows. The update introduces new integrations and expanded provider access to allow agents to move from writing code to deploying live products.

The expansion includes several key developments:

• Infrastructure Provisioning: Hermes, an open-source AI agent from Nous Research, can now use Stripe Projects as a skill. This allows the agent to handle infrastructure management while maintaining context across sessions. Additionally, model-agnostic coding agents Factory Droids and Warp have integrated with Projects, allowing developers to access the embedded Projects CLI within their existing agent workflows.
• Expanded Ecosystem: Stripe Projects now connects to 16 new providers, bringing the total to 49 providers. These additions include Metronome for usage-based billing, Wix for storefronts, and ClickHouse for LLM observability.

This expansion enables a new primitive for software engineering: the ability for an agent to spin up a product, implement billing, and monitor model latency and quality without human intervention in a dashboard. The focus is moving from how humans use tools to how tools enable agents to use other tools.

As agents take over the heavy lifting of provisioning and service connection, the primary challenge for developers will shift from manual integration to the implementation of guardrails.

How will you manage the security implications of an agent that has the power to provision your entire stack?