The evolution of agentic surfaces: building with Claude Managed Agents

2026-06-10 · Source: Claude Blog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Cloud Computing & IT Infrastructure · Depth: Intermediate, long

Summary

Claude Managed Agents, Anthropic's suite of composable APIs, enables teams to reliably deploy production-grade AI agents in production environments at scale. It addresses common challenges like infrastructure overhead, security, and scaling that often prevent prototypes from reaching production. Evolving from simpler "tokens in, tokens out" APIs and the Claude Agent SDK, Managed Agents decouples the agent's "brain" (harness) from its "hands" (execution sandbox). This architecture enhances security by isolating credentials, reduces latency by 60% (p50) and over 90% (p95) in time-to-first-token, and provides persistent sessions with built-in observability and memory features like Dreaming. It supports both Anthropic-managed and self-hosted cloud containers, including MCP tunnels for private networks, with customers like Notion, Rakuten, and Sentry already using it.

Key takeaway

For MLOps Engineers or teams building production-grade AI agents, if you are struggling with infrastructure overhead, security, or performance bottlenecks, Claude Managed Agents offers a robust solution. Its decoupled architecture enhances security by isolating credentials and significantly reduces latency. You can focus on agent context and domain expertise, offloading complex harness tuning and scaling to Anthropic. Consider exploring its features like persistent sessions, built-in observability, and flexible deployment options to accelerate your agent's path to production.

Key insights

Claude Managed Agents decouples agent logic from execution, streamlining production deployment and enhancing security, performance, and reliability.

Principles

• Infrastructure separates prototype from production.
• Decouple agent brain from execution sandbox.
• Harnesses must evolve with model intelligence.

Method

Claude Managed Agents defines agents (config), environments (execution context), and sessions (runs). Sessions persist event history and sandbox state server-side.

In practice

• Use Vaults for secure credential management.
• Delegate orchestration to Anthropic-managed containers.
• Utilize Dreaming for agent memory refinement.

Topics

• Claude Managed Agents
• AI Agent Deployment
• Agent Orchestration
• Production Infrastructure
• Security Isolation
• Low-Latency Inference

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Engineer, Machine Learning Engineer, MLOps Engineer

Related on AIssential

• Claude Managed Agents: get to production 10x faster — Claude Managed Agents simplify AI agent deployment by abstracting infrastructure, accelerating development from months to days.
• Anthropic's Claude Managed Agents Just KILLED 1000+ AI Startups Overnight — Anthropic's Claude managed agents integrate core infrastructure, simplifying AI agent development and shifting focus to outcomes.
• New in Claude Managed Agents: run agents on a schedule and store environment variables in vaults — Claude Managed Agents now automate tasks via scheduling and securely integrate external tools using vaulted environment variables.
• Claude Managed Agents Just Automated EVERY Job! AI Agent OS! — Anthropic's Claude Managed Agents platform simplifies AI agent deployment for complex, long-running tasks.
• Scaling Managed Agents: Decoupling the brain from the hands — Decoupling AI agent components into stable interfaces enables adaptability to evolving models and scalable, resilient operations.
• How to Use Claude Managed Agents? — Anthropic's Claude Managed Agents simplifies AI agent deployment by providing a hosted, stateful, and secure execution environment.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Claude Blog.