Home Get started Platforms Docs GitHub
Italiano English
Vibe coding · AI agents · GitHub Issues

Bugs in chat.
Issues on
GitHub.

With LLMs and fast iteration you lose track of bugs, tasks, and ideas buried in conversation. Issuebeam wires any AI agent (Cursor, Claude Code, Copilot, …) to the official tracker on Windows, macOS, and Linux: the agent runs python scripts/github_issue.py and creates real issues — no gh CLI, no PowerShell scripts.

CLI
Stdlib Python only
Agents
Cursor · Claude · Copilot
OS
Win · macOS · Linux
MIT
Open source
issuebeam — ai agent
# user in chat
"Open an issue for the Safari login form bug"
$ python scripts/github_issue.py create "[Bug] Login Safari redirect loop" --labels bug,priority-high
CREATA #12: [Bug] Login Safari redirect loop
$ python scripts/github_issue.py list
#12 open · bug · priority-high · area-frontend
→ team-visible backlog, not buried in chat
Try with a real bug

01. How it works

From chat to official backlog

Describe work in natural language. The agent checks duplicates, creates the issue with the right labels, and references the number in commits. GitHub Issues becomes the source of truth.

You chat

"Track this bug…", "Open an issue for…", "List open issues". No need to memorize CLI syntax.

Agent runs

Instructions in AGENTS.md (plus Cursor, Claude, and Copilot files) force the agent to run the Python script — not ask you to paste commands.

Issue on GitHub

Shared backlog, labels, web templates, auto-close with Fixes #N in PRs.

02. FAQ

Before you clone

What is Issuebeam?
Issuebeam is a lightweight skeleton that connects AI coding agents to GitHub Issues on Windows, macOS, and Linux. The agent runs a Python CLI to create, comment, and close issues — no gh CLI, no PowerShell scripts. Works with Cursor, Claude Code, Copilot, and any agent that can read repo instructions and run shell commands.
Is it Cursor-only?
No. Cursor is the best-tested path (.cursor/rules/github-issues.mdc), but the same CLI works with Claude Code (CLAUDE.md + AGENTS.md), GitHub Copilot (.github/copilot-instructions.md), Windsurf, Cline, Gemini CLI, Codex CLI, and manual use. See the Platforms page or the MkDocs guide at /docs/platforms/overview/.
Do I need the gh CLI?
No. Issuebeam uses a stdlib-only Python script (urllib) that talks directly to the GitHub REST API. You only need Python and a Personal Access Token with Issues read/write.
How does the agent know to use it?
Copy the skeleton into your repo (or use adopt.py). Each platform reads an instruction file: AGENTS.md (universal), .cursor/rules/ for Cursor, CLAUDE.md for Claude Code, .github/copilot-instructions.md for Copilot. The rule tells the agent to run python scripts/github_issue.py directly.
Can I adopt it into an existing project?
Yes. Run python scripts/adopt.py --target ../my-repo --repo myorg/my-app. It copies the CLI, labels, issue templates, and agent instruction files for multiple platforms, then writes tracker/github_repo with your slug.
Where does the token live?
Set GITHUB_TOKEN as an environment variable (Windows, macOS, Linux). Alternatives: .env (gitignored) or .secrets/github_token. On Windows, the CLI can also read user env vars from the registry if the IDE terminal does not inherit them. Never commit tokens or paste them in chat.

Full guide: MkDocs · Platforms

Ready to beam issues?

Clone the repo, configure token and slug, apply labels. Then open your preferred AI agent and report a real bug — watch it land on GitHub.