Anthropic Launches Claude Managed Agents to Accelerate AI Production by Ten Times

Accelerating the Agentic Era

Anthropic announced the public beta of Claude Managed Agents on April 8, 2026. This new suite of composable APIs simplifies the creation of autonomous AI systems by handling the underlying infrastructure, including secure sandboxing and state management. Developers can now move from initial prototype to full production roughly ten times faster than previous methods. The platform addresses common hurdles of AI deployment such as credential management and execution tracing. By running agents on Anthropic’s hosted infrastructure, teams no longer need to build bespoke orchestration harnesses or sandboxed code environments. Instead, they define specific tasks and tools while the system manages context and error recovery automatically.

Advanced Performance and Features

Key features include long-running sessions that operate for hours and a research preview for multi-agent coordination. Internal testing shows that these managed loops improve success rates by up to 10 points on the most difficult tasks. The architecture is designed to be future-proof, ensuring that agent loops remain stable even as underlying models are upgraded. Several major platforms are already utilizing the technology to enhance their workflows. Notion uses the system for custom agents that handle complex coding and document generation within workspaces. Meanwhile, Sentry has integrated the tool to allow their debugging agent to write patches and open pull requests autonomously.

"Managed Agents not only allowed us to build the initial integration in weeks instead of months, but has also eliminated the ongoing operational overhead of maintaining bespoke agent infrastructure," stated Indragie Karunaratne, Senior Director of Engineering at Sentry.

Enterprise Adoption and Pricing

Industry leaders like Atlassian and Rakuten have also adopted the platform to scale AI teammates across sales and finance. Sanchan Saxena, Head of Product at Atlassian, noted that the system allows their engineers to focus on features rather than infrastructure. This shift significantly reduces the operational overhead traditionally associated with maintaining secure agent runtimes. Managed Agents is currently available on the Claude Platform under a consumption-based pricing model. In addition to standard token rates, users pay a flat fee of $0.08 per session-hour for active runtime. Developers can get started via the Claude Console or the new command line interface.