Claude by Anthropic Can Now Coordinate Up to 1,000 AI Agents Simultaneously via Dynamic Workflows

Meta description: Anthropic brings dynamic multi-agent workflows to Claude Managed Agents, letting a lead agent plan, split tasks, and merge results across up to 1,000 agents.

Tags: Claude Managed Agents, Anthropic, multi-agent orchestration, dynamic workflows

Featured image: abstract technology-style network of connected nodes representing Claude Managed Agents dynamic workflows, no people, no logos, no text.

Anthropic is expanding Claude Managed Agents by adding dynamic workflows, a change that moves the platform from isolated managed agents toward coordinated groups of agents. The update matters because it gives Claude a built-in structure for splitting complex work, assigning it to multiple agents, and combining their outputs. For developers, this makes advanced automation easier to assemble. For operators, it also makes cost and coordination overhead more visible.

Background: A Managed Agent Layer Was Already in Place

Claude Managed Agents already existed as infrastructure for running agents inside Anthropic's platform. The underlying managed agent setup had been available for a while. What was missing was a first-class way for several agents to operate as part of one orchestrated process.

Dynamic workflows are the new layer on top of that existing foundation. Orchestration, in this context, means the automatic planning, distribution, and merging of work across agents instead of requiring a human to chain every step manually.

This timing also sits against a skeptical view of multi-agent systems. A senior OpenAI engineer recently called agent swarms a massive waste of tokens. Anthropic's move suggests the company sees coordination as useful enough to make it a native feature of Claude Managed Agents.

Development Process: How the New Workflow Structure Operates

The dynamic workflow model centers on a lead agent. That lead agent creates a plan, distributes tasks to sub-agents, and merges results when the work is done. A sub-agent acts as a worker instance that receives a portion of the task and returns an output.

The system can use up to 1,000 agents in parallel per execution. That figure is a key number in the announcement. Parallelism means multiple agents can handle parts of a job at the same time, which may help with long tasks that can be divided cleanly. It also introduces coordination cost, because the lead agent must bring the pieces back together.

The activation path is specific. Users select the `multiagent_20261001` agent type to enable the feature. Anthropic also points to documentation or to running `/claude-api managed-agents-onboard` in Claude Code as the way to get started.

Latest Progress: Test Results and the Cost Warning

A test described by Anthropic placed 70 bugs in a 116,000-line codebase. A single agent caught between 14 and 27 bugs per run. The dynamic workflow consistently hit 66. That result explains the pitch for orchestration: when a task can be divided, a wider search may find more issues than one agent can cover alone.

The test does not settle whether those gains hold across task types. Anthropic says the pattern still needs to be tested against real workloads, and users should evaluate whether the workflow structure fits their own applications.

The cost caveat is equally important. Anthropic says these workflows can burn through "a lot of tokens," and it recommends starting small. A token is the basic unit of text processed by a model, so more agents can mean more prompt and response traffic even when each individual step is short.

Current Stage: Early Orchestration with Measurable Tradeoffs

Claude Managed Agents is now at a stage where multi-agent coordination is available as a defined feature rather than a pattern users must build from scratch. The timeline is simple: managed agent infrastructure first, dynamic workflows second, and a concrete activation path through the `multiagent_20261001` agent type.

What to watch next is how users apply the feature beyond the reported codebase test, and whether Anthropic adds clearer guidance on cost controls, task routing, and result merging. The update expands what Claude Managed Agents can do, but it also makes token consumption and workflow design more visible parts of the decision.