Claude Managed Agents (Flows)

Flows in Claude Managed Agents is an Anthropic API feature for running dynamically planned multi-agent workflows. A user supplies a goal, after which Claude creates and executes a topology of up to 1,000 agents, dynamically starting and stopping sub-agents across workflow phases. Demonstrated templates include a stock watchlist scanner, Bug Hunter for pull-request review, and Explainer Studio for video generation. The feature includes per-phase compute budgets, interrupts, automatic operation modes, credential vaults, and network lockdown controls for operating agent workflows at scale.

Claude Managed Agents is Anthropic's managed agent infrastructure: a configurable agent harness for running Claude on long-running and asynchronous tasks. An agent combines a model, system prompt, tools, MCP servers, and skills; an environment specifies an Anthropic-managed cloud sandbox or a self-hosted sandbox; and a session runs the agent and produces persisted events. Applications send user messages as events, while Claude runs tools and streams responses through server-sent events; sessions can be steered or interrupted, and their event history can be retrieved. The service provides managed tool execution and sandbox infrastructure, including shell commands, file operations, web search and fetch, and MCP connections. It includes prompt caching, compaction, persistent filesystems, conversation history, and server-side storage, so users do not need to build their own agent loop, sandbox, or tool-execution layer. The documentation also describes a design that separates the agent's harness and model from its sandboxes and tools, with credentials handled outside the sandbox through an egress proxy. Claude Managed Agents is in beta and requires the managed-agents-2026-04-01 beta header. Stateful sessions are not currently eligible for Zero Data Retention or HIPAA Business Associate Agreement coverage.

Claude Platform Console is Anthropic's web console for building agents and applications with Claude models and managed agent infrastructure. Its guided quick-start flow lets users create an environment, agent, and session without writing code. The managed-agent flow separates the session event log, the agent's harness and Claude-based “brain,” and sandboxed tools or “hands”; credentials are injected through an egress proxy rather than exposed to the sandbox.

Claude CoWork is an AI agent platform for building and running task-specific skills and agents. The documented workflow is to select a repeated process, teach the agent with real cases, package the process as a skill, test it against edge cases and real examples, then deploy it with an assigned owner and reviewer. In the financial-planning example, agents draft cash-flow analyses, insurance reviews and tax reviews while a certified financial planner reviews the results before delivery. Claude is available through web, desktop and mobile applications and can create files, run code, search the web and connect to apps and tools.

Claude Console is Anthropic’s web console for building applications and managed agents with Claude models. It provides API-key creation, a natural-language quick start for creating an agent, deployment search and triggering, and live session transcripts with a debug event stream that includes model latency. The managed-agent workflows described for it include scheduled deployments, fan-out across sessions with custom tools, workspaces for tenancy and access, and a research-preview memory process called Dreaming that reviews past sessions to improve a memory store.