GitHub has released a public preview of the dynamic workflows feature for controlling agents via code
The dynamic workflows feature is available in public preview in Copilot CLI, GitHub Copilot app, and GitHub Copilot SDK on all GitHub Copilot plans. It allows defining sequential and concurrent steps and agent involvement in code.
GitHub has made the dynamic workflows feature available in public preview in Copilot CLI, GitHub Copilot app, and GitHub Copilot SDK. It is available on all GitHub Copilot plans, and its form may change.
A dynamic workflow is a program that combines automated steps with the work of one or more agents. The code determines the order and concurrency of steps, when to involve agents, and how to use their results. Agents handle parts that require analysis or judgment. The program runs inside a GitHub Copilot extension and has access to the extension API. A user can write it themselves or have GitHub Copilot generate it.
GitHub cites service incident investigation as an example: collecting logs and telemetry, concurrent analysis of different systems by independent agents, and combining the results into a timeline and root-cause report. Unlike the /fleet command, in which GitHub Copilot delegates and coordinates concurrent work, here the procedure is determined by code. The feature is intended for reusable processes or tasks with clear phases, checks, or constraints.
Why it matters
An individual can convert a recurring task into a program with explicitly defined steps and agent involvement. For enterprise teams, the feature allows unifying, for example, the incident investigation procedure and combining the results of concurrent analyses into a joint output. The source does not document any specific time or cost savings.
Two audiences, two different impacts
What this means
For individuals
For recurring tasks, you can specify in code when the agent should step in and how its result is used in subsequent steps.
For a business
A team investigating incidents can embed a common procedure for data collection, analysis of individual systems, and compilation of a root-cause report into a reusable program.
ProcessesCheck the original
Event sources
clearly official source · 1 publisher, 0 independent. We count feeds from the same owner only once.