Overview¶
The problem¶
With LLMs and fast iteration:
- Bugs discussed in chat are forgotten
- Tasks stay "mental" but never appear in the team backlog
- Scattered markdown files (
TODO.md,bugs.txt) multiply - Nobody knows what is open, in progress, or closed
The solution¶
Vibeissue connects AI agent chat to GitHub Issues with:
pip install vibeissue— CLI on PyPI (vibeissue,vibeissue-adopt)- Agent instructions — files the LLM reads (
AGENTS.md, Cursor rule, Copilot, Claude, …) - GitHub templates — web forms for browser-based issue creation
- Skeleton in your repo — labels, tracker config, and rules via adopt
GitHub Issues becomes the source of truth. Local markdown stays for detailed plans and archives — not operational status.
Architecture¶
Requirements for any platform:
- Python 3 (Windows, macOS, Linux) +
pip install vibeissue GITHUB_TOKENwith Issues read/write- Repo slug in
tracker/github_repoorGITHUB_REPO - Agent instructions: run the CLI, don't delegate to the user
Quick path (CLI only)¶
pip install vibeissue- GitHub token + repo slug in
.env vibeissue list
Optional — full path (AI agents)¶
- Adopt into a project (
vibeissue-adopt) - Open the agent and say "Open an issue for…"
- Choose your platform