Ghar: An Automated Agent Routine Framework Based on GitHub Workflows

Ghar (GitHub Agent Routines) is an open-source project by tbrandenburg (released on GitHub on 2026-06-08, link: https://github.com/tbrandenburg/ghar) that encapsulates AI agent capabilities into directly runnable GitHub Actions workflows and shell scripts. It allows developers to implement advanced functions like automated task processing, Issue auto-fix, and multi-agent collaboration in code repositories without complex configurations.

In modern software development, automation is key to efficiency, but traditional CI/CD tools are limited to fixed processes (build, test, deploy) and struggle with complex scenarios requiring intelligent decisions. Ghar fills this gap by integrating LLM-driven agent capabilities into GitHub workflows, making AI automation accessible to ordinary developers.

Ghar follows the "reusable agent routines" concept with a layered architecture. It includes multiple GitHub Actions workflows:

1. Issue processing (e.g., ghar-issue-opened.yml for classification, ghar-complete-issue-resolution.yml for auto-fix)
2. Multi-agent collaboration (e.g., ghar-multi-agent-tdd-issue-resolution.yml for TDD-based teamwork)
3. Daily routines (ghar-daily-routine.yml for dependency checks, code quality scans)
4. Core execution (core-opencode-run.yml for OpenCode integration) Supporting tools: Makefile.ghar (unified commands), scripts directory (auxiliary functions), platform-specific configs (.claude/.opencode), and Git hooks (.githooks)

• Event-driven: Uses GitHub Actions events (new Issue, PR, cron tasks) to trigger agent routines
• Multi-agent collaboration: In TDD workflow, agents take roles like analysis (understand problem), test (write cases), implementation (fix code), validation (run tests)
• Security: Isolated sandbox environment and GitHub permission control to limit sensitive resource access

• Open-source maintenance: Auto handle common Issues (docs errors, simple bugs) to save maintainers' time
• Enterprise automation: Code standard checks, dependency security scans, automated docs generation
• Personal dev: Auto reply to common questions, generate PR descriptions, sync code across platforms

To use Ghar:

1. Copy the .github/workflows directory from Ghar to your repo
2. Configure environment variables and keys as needed
3. Refer to AGENTS.md for detailed agent configs For local development: Use Makefile.ghar commands to test workflows without GitHub Actions

Ghar represents a new paradigm: AI agents as cloud services via GitHub Actions. Its advantages: zero operation and maintenance cost (uses GitHub infrastructure), version control (workflows in Git), easy extension (YAML configs), community sharing. Future: As LLM capabilities improve, Ghar will play a bigger role in software development, moving toward "code as agent"

Ghar seamlessly integrates AI agent capabilities with DevOps toolchains. It automates repetitive tasks, allowing developers to focus on creative work. For teams exploring AI automation, Ghar provides a low-threshold, high-return starting point