Anthropic has launched Claude Managed Agents, a new set of APIs designed to help developers build and deploy cloud-based AI agents faster by handling infrastructure, orchestration, and runtime management.
The product, now available in public beta on the Claude Platform, is positioned as a way to reduce the time required to move agents from prototype to production. Anthropic said the system combines a prebuilt agent harness with managed infrastructure so teams can focus on defining tasks and workflows rather than building underlying systems.
Managed Agents supports core capabilities required for production use, including secure sandboxed execution, authentication, permission controls, and tool integration. The platform also enables long-running sessions that can operate autonomously for extended periods, along with orchestration features that allow multiple agents to coordinate on complex tasks.
Anthropic said the system is designed to handle operational challenges that typically slow deployment, such as state management, error recovery, and tracing. Developers can monitor agent behavior through built-in tools that track decisions, tool calls, and failure modes.
The company said internal testing showed improvements in task success rates compared with standard prompting approaches, particularly on more complex workflows. Managed Agents also supports both autonomous execution and traditional prompt-based interactions, depending on how developers want to structure control.
Early adopters are already using the system across a range of use cases. Companies including Notion, Rakuten, Asana, and Sentry are deploying agents for tasks such as writing code, processing documents, managing workflows, and generating deliverables. In several cases, Anthropic said teams were able to launch these systems in weeks instead of months.
Pricing is based on usage, with standard Claude token rates plus an additional fee for runtime. The service is accessible through the Claude Platform, along with tools such as the Claude Console and command-line interface.
The launch reflects a broader shift in AI development toward managed systems that abstract away infrastructure complexity, as companies look to accelerate deployment of agent-based applications.
About this article: This article was generated with AI assistance and reviewed by our editorial team to ensure it follows our editorial standards for accuracy and independence. We maintain strict fact-checking protocols and cite all sources.
Word count: 348Reading time: 0 minutes
Explore More AI Resources
Continue with high-value guides related to this topic.
