GitHub’s video, presented by the GitHub team, unveils the new ‘agentic workflows’ feature, empowering users to automate tasks via markdown descriptions that AI agents—like GitHub Copilot—can execute as GitHub Actions workflows.

GitHub Introduces Agentic Workflows: AI Agents for Automation

GitHub has announced support for agentic workflows, representing a new way to automate developer tasks directly through AI-powered agents.

What Are Agentic Workflows?

  • User Experience: Users describe the task or workflow in a simple markdown file.
  • Automation: This markdown description is compiled into a GitHub Actions workflow, automating tasks that previously required manual configuration.
  • AI Agent Integration: Any compatible AI agent—such as GitHub Copilot, Claude Code, or OpenAI Codex—can interpret and execute these workflows, enabling intelligent automation on the GitHub platform.

Key Features & Benefits

  • Democratized Automation: Lower barrier for developers by leveraging markdown for workflow instructions rather than complex YAML or scripting.
  • Flexible Agent Support: Works with a range of AI agents, not limited to GitHub Copilot, expanding automation and assistant options.
  • Integration with Existing GitHub Ecosystem: Runs through GitHub Actions, supporting existing CI/CD and automation practices.

Example Flow

  1. Developer describes a task in markdown (e.g., CI build, deployment steps).
  2. The agent processes and compiles it into a production-ready GitHub Actions workflow.
  3. The workflow is executed, potentially involving multiple agents or fallback strategies.

Supported AI Agents

  • GitHub Copilot: Integrated developer assistant for workflow creation and execution.
  • Claude Code: Third-party AI code agent.
  • OpenAI Codex: Language model for translating natural language to code.

More Information

Stay tuned for updates as this feature becomes more widely adopted and further usage patterns and best practices emerge.