Claude agents can now write their own multi-agent workflows, with the server running each one in the background
Synopsis
Anthropic added dynamic workflows in beta to Claude Managed Agents on the Claude Platform: for work with many pieces, such as reviewing hundreds of documents, an agent can write a program that runs many agents in phases and combines their results, and the server runs it in the background as a workflow run. It is turned on by setting the agent's multiagent field to multiagent_20261001 with workflows enabled; the user can tell the agent in its system prompt when to start a run and follow each run through workflow_run.* events on the session's event stream.
Interpretation
An agent in Claude Managed Agents can now write a workflow itself for work with many pieces, such as reviewing hundreds of documents. Multi-agent orchestration was already available in beta; now the agent writes that orchestration program itself during the task rather than the user writing it in advance. The capability is in beta, requires the managed-agents-2026-04-01 beta header, and is enabled by setting the agent's multiagent field to multiagent_20261001 with workflows enabled.
A workflow is a program that runs many agents in phases and combines their results, and the server runs it in the background as a workflow run. It is turned on by writing the multiagent_20261001 type into the agent's multiagent field with workflows enabled, rather than by hand-writing the orchestration. The user can tell the agent in its system prompt when to start a run and follow each run through workflow_run.* events on the session's event stream.
Perspective
The capability targets Claude Managed Agents users and is in beta, requiring the managed-agents-2026-04-01 beta header; the release notes' example setting is work with many pieces, such as reviewing hundreds of documents.
The release notes give no effectiveness data on specific tasks and do not describe how a workflow run handles failure or an error in one phase.