Anthropic Launches Platform for AI Agent Deployment

Anthropic unveils Claude Managed Agents for complex AI tasks.

AI startup Anthropic has introduced an environment for executing complex and prolonged agent tasks—Claude Managed Agents.

New on the Engineering Blog:

Building Managed Agents—our hosted service for long-running agents—meant solving an old problem in computing: how to design a system for “programs as yet unthought of.”

Read more: https://t.co/YYaEub2QGV

— Anthropic (@AnthropicAI) April 8, 2026

The solution is a hosting service on the Claude platform, enabling the management of long-term agents through a set of basic interfaces.

The system can sequentially execute tasks, use tools, run code, edit files, interact with external services, and continue functioning even after failures.

The Problem

Previously, developers faced challenges when deploying AI agents:

• digital assistants would forget context, necessitating a reset;
• the model would prematurely end a task, requiring special “workarounds”;
• the neural network struggled with lengthy assignments.

For instance, Claude Sonnet 4.5 could prematurely conclude a task as the context window limit approached.

The Solution

The cloud-based tool Claude Managed Agents handles all background processes, including memory management and error recovery.

The main architectural concept of the new solution is the division of the agent into independent components. Previously, everything operated in one place:

• the model itself and the logic of its invocation;
• code execution;
• memory and session states;
• accesses and tokens.

Anthropic divided the system into three main blocks:

1. Session—a log of all actions and events. It acts as the agent’s long-term memory, recording the history of steps and bypassing context length limitations.
2. Control logic—this “conductor” calls Claude, decides which context to provide the model, and manages the work cycle.
3. Execution environment—a place where LLM can run code, edit files, and interact with the environment.

This structure allows each part to function autonomously—a local failure in one of the processes no longer leads to the closure of the entire session.

Anthropic positions Claude Managed Agents as a resilient infrastructure layer that will outlast the evolution of models. It is akin to an operating system for AI agents.

Previously, the company developed a new LLM Claude Mythos, but declined to release it publicly due to high security risks.