Anthropic’s new product aims to handle the hard part of building AI agents

At a glance:

• Anthropic launches Claude Managed Agents to simplify AI agent deployment.
• Revenue exceeds $30 billion, three times December 2025 levels.
• The service provides sandboxed, scalable infrastructure for autonomous agents.

How Claude Managed Agents Work

Claude Managed Agents introduces an agent harness that bundles the software infrastructure needed for an AI model to act autonomously. The harness includes a memory layer, tool‑calling wrappers and a secure execution environment, allowing the model to plan, retrieve information and invoke external services without manual wiring. Developers can launch agents that run for hours in the cloud, monitor the activity of other agents, and adjust permissions that grant access to specific APIs or databases. This level of control makes it possible to build complex workflows such as data‑pipeline orchestration or real‑time customer support triage. In a WIRED demo, Notion showed how a Claude Managed Agent could sequentially complete onboarding tasks, while the developer observed the agent’s actions on a dashboard hosted on the Claude Platform. The visual feedback helps teams verify that the agent is following the intended logic.

Market Impact and Competitive Landscape

Anthropic announced that its annualized recurring revenue has topped $30 billion, a figure roughly three times larger than the December 2025 baseline. This surge reflects strong demand for the Claude Platform API, which underpins both the new Managed Agents service and other enterprise offerings. The company’s push mirrors OpenAI’s own Frontier agent platform, creating a head‑to‑head race among the leading AI firms to provide turnkey solutions for large‑scale automation. Both firms are positioning their stacks as the backbone of next‑generation enterprise software. Wall Street analysts have begun to question whether traditional software‑as‑a‑service models will survive the rise of self‑sufficient AI agents, noting that Anthropic’s rapid feature rollout could erode margins for legacy SaaS providers.

Enterprise Adoption and Use Cases

Beyond Notion, a variety of enterprises are experimenting with Claude Managed Agents to automate repetitive processes such as invoice reconciliation, ticket triage and inventory updates. The service’s plug‑and‑play nature lets non‑technical teams prototype agents without hiring specialized engineers. Because the platform handles scaling, monitoring and permission management, businesses can offload long task lists to a single agent that executes steps one after another, freeing human workers to focus on strategic decisions. Katelyn Lesse, head of engineering for the Claude Platform, emphasized that many customers previously needed entire teams to build the distributed systems required for agent deployment; now those teams can redirect effort toward product differentiation.

Technical Architecture and Security

Each managed agent operates inside a sandboxed environment that isolates its code and data from the host system, preventing unintended interactions with other workloads. When an agent needs to start a new project, the platform provisions a secure container and spins up the necessary dependencies automatically. Permission toggles let administrators specify which tools an agent may call, such as accessing a proprietary database or invoking a third‑party API, thereby reducing the attack surface while preserving flexibility. Lesse explained that the underlying distributed‑systems engineering — load balancing, fault tolerance and state replication — remains a non‑trivial challenge, but Anthropic abstracts much of this complexity away from the end user.

Outlook and Future Developments

The rollout schedule indicates that Claude Managed Agents will be generally available later this year, with plans to expand support to additional cloud providers and to integrate with popular development frameworks. Future updates are expected to add multimodal capabilities, allowing agents to process images, audio and video in addition to text, which could open new use cases in content moderation and multimedia analysis. If the trajectory holds, enterprises may eventually rely on fleets of autonomous agents to manage entire workflow pipelines, shifting the role of human workers toward oversight, strategy and creative problem‑solving.