GitHub Next has announced the release of Agentic Workflows in technical preview phase. This is an open source project that explores what security measures are necessary for AI agents to run continuously within repositories.
The project are automated repository workflows and based on intents running on GitHub Actions, written in simple Markdown and executed with coding agents. They focus on recurring tasks that require a lot of thinking, such as maintenance, issue triage, CI failure analysis, documentation drift, and code hygiene. Workflows run on a schedule within existing permission models, with explicit limits built in.
The team is aggressively testing its own product. They intentionally built Agentic Workflows in Go without knowing Go, and then created a “go-fan” workflow that served as a daily, continuous feedback loop that improved both their code and the system at the same time. They are designed for people who work on GitHub, from individuals automating a single repository to teams operating at enterprise or open source scale.
Agentic workflows are based on broader research the company is conducting around continuous AI. At the same time, they are presented as one of the many options that GitHub is exploring, along with its commitment to help repository managers to better manage the large volumes of contributions they receive in this age of AI.
How Agentic Workflows works
The concept behind GitHub Agentic Workflows is simple: the user describes the results they want in simple Markdown, adds it as an automated workflow to their repository, and executes it using an encoding agent in GitHub Actions. This integrates the power of encryption agents into repository automation.
Agentic Workflows run as standard GitHub Actions workflowswith additional security measures for sandboxing, permissions, monitoring, and review. When running, they can use different coding agent engines, such as Copilot CLI, Claude Code, or OpenAI Codex, depending on the configuration.
Using GitHub Agentic workflows makes it possible completely new categories of repository automation and software engineeringso it fits perfectly with the way development teams already work on GitHub. All of them would be difficult or impossible to implement with traditional YAML workflows alone, including:
- Continuous triage– Automatically summarize, tag, and address new issues.
- Continuous documentation– Keep README files and documentation aligned with code changes.
- Continuous code simplification– Repeatedly identify code improvements and open pull requests for them.
- Continuous testing improvement– Evaluate test coverage and add high-value tests.
- Continuous quality hygiene– Proactively investigate CI failures and propose specific solutions.
- Continuous reporting– Create regular reports on repository status, activity, and trends.
These are just some examples of repository automations that demonstrate the power of GitHub Agentic workflows or what repository service providers call Continuous AI: the integration of AI into the software development lifecycle (SDLC), which improves automation and collaboration similar to continuous integration and continuous deployment practices.
GitHub Agentic and Continuous AI workflows are designed to complement existing CI/CD, rather than replace it. They do not replace build, test, or release pipelines, and their use cases do not overlap greatly with deterministic CI/CD workflows. Agentic workflows run in GitHub Actions because that’s where GitHub provides the necessary infrastructure for permissions, logging, auditing, sandboxed execution, and extensive repository context.
