Git-issues – AI agent-first task management for Git repos
Issues as versioned Markdown files means you can branch, revert, and diff task history with git.

Issues as Markdown files in your repo means they finally travel with your branches.
Solo developers and small teams wanting offline-first, git-synced issue tracking
GitHub Issues · Linear · Todo.txt
- my very own (probably flawed) workflow sometimes changes planned features along the way which tends to get source code and feature descriptions out of sync - I wanted Claude Code (or Codex) to be faster when accessing the planned tasks - have everything checked out that's needed for work to be able to work offline (sometimes happens on a boat etc.)
So, in short:
Issues stored as YAML-frontmatter Markdown files in .issues/, version-controlled alongside your code. Zero infrastructure: one Go binary, no server, no database, no accounts.
Built for two workflows: - Human: issues list, issues show, issues board (interactive TUI) - Agent: issues next → issues claim <id> → issues done <id>. Auto-generates .agent.md context for Claude Code and other agents.
Relations (blocks, depends-on, related-to, duplicates) are bidirectionally synced. Every change auto-git add-ed. Issues travel with branches, git bisect shows state at any commit.
go install github.com/steviee/git-issues@latest
What do you think?
Issues as versioned Markdown files means you can branch, revert, and diff task history with git.
Git-native issues AI agents can manage, but GitHub Issues + MCP already solve this better.
Git-native issues as markdown files, but GitHub Issues, Gitea, and Fossil already solve this.
Issues as Git commits, finally solving Linus's 2007 proposal after 17 years.
Git-based CMS for JSON and YAML files when TinaCMS and Decap already exist.
Git-native event sourcing for issues beats Jira's bloat for terminal purists.