Stop Giving AI Agents Your Master Keys

Zalando open-sourced a tool that lets AI agents securely access company systems without getting a user's full permissions, a key enterprise challenge.

Zalando’s internal incubator has released a new open-source tool designed to solve a critical security challenge for companies using AI agents. The project, called the Agentic Identity Broker (AIB), provides a secure way for a user to delegate tasks to an AI agent without handing over their complete set of credentials. As announced on Hacker News by one of its co-authors, the MIT-licensed tool is already used in production at Zalando, lending it significant credibility. The core problem AIB addresses is that as AI agents become more capable, they need access to internal company services to perform useful work, but giving them a user's full access tokens creates a massive security risk. The AIB acts as a middleman to grant limited, temporary, and revocable permissions, much like giving a valet a key that only starts the car but doesn't open the trunk.

The Agentic Identity Broker functions as a specialized OAuth2 broker, a standard and widely understood protocol for authorization. When a user wants an AI agent to perform an action, such as accessing a calendar or a database, the user first authenticates with the AIB. Instead of passing the user's powerful, long-lived credentials to the agent, the AIB generates a new, narrowly-scoped token specifically for that agent to complete that single task. This new token has a short lifespan and is restricted to only the permissions necessary for the job. The agent then uses this temporary token to interact with the target service. This mechanism ensures that if the agent or its session is compromised, the potential damage is strictly contained, and the access can be instantly revoked without affecting the user's primary credentials.

This release arrives as the industry grapples with the practicalities of moving AI from simple chatbots to autonomous agents integrated into core business workflows. Many companies are eager to empower agents to automate complex processes, but the security implications have been a major roadblock. Current methods are often insecure workarounds, such as developers storing sensitive API keys directly in prompts or configuration files, which is a notoriously poor practice. The AIB represents a significant step toward establishing mature, secure infrastructure for the growing world of AI agents. It treats agents not as simple scripts, but as distinct actors within a system that require their own robust identity and access management (IAM) framework. This approach shifts the paradigm from ad-hoc access hacks to a structured, auditable, and secure architecture.

For CTOs, security teams, and developers building with AI, the Agentic Identity Broker offers a practical, production-tested blueprint for a problem many are currently trying to solve from scratch. Adopting an open-source solution like AIB can save significant development time and, more importantly, help avoid common security pitfalls associated with managing agent permissions. Its release signals a broader trend toward standardizing the infrastructure needed to safely deploy powerful AI agents within corporate environments. As more companies look to automate workflows, we can expect to see a greater focus on developing and adopting specialized security tools like the AIB, making the secure integration of AI a more achievable goal for organizations of all sizes. Teams should evaluate this and similar tools as a foundational component of their enterprise AI strategy.

For developers and security teams, this provides a standardized, open-source pattern for managing AI agent permissions. It solves the critical problem of how to grant agents scoped, temporary access to APIs without exposing a user's long-lived, high-privilege credentials, enabling safer, more powerful integrations.

The Agentic Identity Broker can accelerate the adoption of AI agents in the enterprise by mitigating a major security risk. This allows companies to automate more complex workflows and integrate AI into core business processes with greater confidence, unlocking significant productivity gains.