Your Newest Platform User is an AI Agent — Build it a Golden Path

The landscape of platform engineering is rapidly changing with the advent of AI agents, which are now being recognized as a primary user of Internal Developer Platforms (IDPs). Traditionally, IDPs have been designed with human developers in mind, focusing on streamlining workflows, providing self-service capabilities, and establishing 'golden paths' for efficient software delivery. However, as of 2026, AI agents have entered this ecosystem, demanding a re-evaluation of how platforms are built and managed. Platform engineering teams are increasingly realizing that these AI agents, which operate tirelessly and can process documentation instantly, require the same level of consideration as human developers. This means extending existing concepts like 'golden paths' – the curated, recommended workflows – to accommodate autonomous systems. The goal is to enable AI agents to interact with the platform programmatically, performing tasks such as provisioning services or triggering deployments without direct human intervention. A crucial aspect of this evolution is the shift in the IDP's role. It is no longer just a portal for human interaction but is becoming a sophisticated control plane for AI agents. The Model Context Protocol (MCP) is identified as a key enabler for this transformation, allowing AI agents to discover and invoke platform actions as native tools. This capability facilitates seamless integration of AI into development processes, from scaffolding new services to requesting database access. However, this integration comes with significant governance challenges. Platform engineers must carefully define what capabilities and tools are exposed to AI agents. Instead of granting broad access to raw infrastructure APIs like `kubectl` or entire cloud provider APIs, the focus is on providing a curated set of explicitly scoped actions, such as `scaffold_new_service` or `request_database`. This approach ensures that AI agents operate within defined guardrails, maintaining security, compliance, and control. The article underscores that the core principles of platform control – golden paths, guardrails, safety nets, and human review – are not replaced by AI but become essential frameworks for its safe adoption.