Anthropic has moved beyond model provision with a managed platform that handles infrastructure, security, and governance for enterprise AI agents. The Claude Managed Agents platform entered public beta today, offering businesses a way to deploy production-ready agents without building custom infrastructure.
Developers define tasks, tools, and permissions within a managed environment while the platform handles execution and state management. This includes secure sandboxing for code execution, long-running autonomous sessions that persist through disconnections, and multi-agent coordination for parallelizing complex work.
"Until now, building agents meant spending development cycles on secure infrastructure, state management, permissioning," according to Anthropic's announcement. The company claims its platform enables customers to "go from prototype to launch in days rather than months."
Enterprise governance features address a key barrier to production adoption. The platform provides scoped permissions for accessing real systems, identity management controls, and end-to-end execution tracing built directly into the service.
These tools give enterprises visibility into agent activity while maintaining security boundaries.
Pricing follows a consumption model with standard Claude API token rates plus an additional $0.08 per session-hour for active runtime. When agents perform web searches through the platform, Anthropic charges $10 per 1,000 searches on top of model usage costs.
Internal testing focused on structured file generation showed Managed Agents improving task success rates by up to 10 points over standard prompting loops. The largest performance gains occurred on the most difficult problems where traditional approaches struggled.
Several advanced capabilities remain in research preview requiring special access requests. These include outcomes-based iteration where Claude self-evaluates against success criteria until achieving results, multi-agent coordination systems for spinning up subordinate agents, and persistent memory tooling that retains context across interactions.
The launch positions Anthropic more directly against OpenAI and Microsoft's enterprise AI offerings by providing not just models but complete execution environments. While Claude Code established a developer ecosystem around its coding capabilities earlier this year, Managed Agents represents Anthropic's first infrastructure-as-a-service play for running autonomous agents at scale.
