Anthropic's Claude Code hits 131K GitHub stars: why developers are skipping the IDE

Most AI coding assistants ask you to change how you work. Open a new tab, paste your code, wait for a response and copy it back. The round trip adds friction to every question.

If you work primarily in the terminal, that friction is worse. Claude Code is built for the way you already work. Anthropic's terminal-based coding agent just crossed 131K GitHub stars and 21.2K forks, with 657 commits and 54 contributors as of June 2026.

What I keep coming back to: Anthropic's own GitHub Actions workflows now use Claude Code to triage issues and deduplicate bug reports on the Claude Code repository itself. A company using its own tool in production, on the repo where that tool is actively developed, is a different signal than a star count.

What Happened

Claude Code reached general availability with version 1.0.0 in May 2025 and has been in steady development since. The repository last saw its changelog and feed updated on June 6, 2026.

One thing worth noting: npm installation is now deprecated. Anthropic recommends native installers for macOS, Linux, and Windows, including Homebrew and WinGet. If you evaluated it early and hit setup friction, that path is cleaner now.

The addition of a feed.xml alongside the changelog tells me Anthropic is treating this as a product with a real changelog audience, not just a research artifact. Teams that are actively developing can subscribe to the feed instead of manually monitoring the repo.

Key Features

• Terminal-native agent: Runs in your shell with a single claude command from your project directory. No browser tab, no context switching.
• Natural language git workflows: Handles commits, diffs, and PRs through plain English. Tag @claude directly on GitHub to trigger actions on PRs and issues without opening a terminal session.
• Cross-platform install: Available via curl on macOS and Linux, irm on Windows, Homebrew, and WinGet.
• Plugin system: A plugins directory ships with the repo, supporting custom commands and agents that extend the tool's default behavior. The plugins directory was updated two weeks ago, which tells me this surface area is actively being worked on.
• IDE and GitHub integration: Works alongside VS Code via a dedicated extension, with devcontainer support and direct GitHub PR/issue integration via @claude mentions.
• Workload identity federation in CI: Recent GitHub Actions updates have added workload identity federation for Claude auth in CI workflows, paving the way for teams running Claude Code in automated pipelines without storing credentials.

Why It Matters

The self-hosting detail is the one I keep pointing teams toward. Anthropic's GitHub Actions workflows use Claude Code to triage issues on its own repository. That's a team trusting the tool with real workflow automation on a production codebase, not a demo environment.

For teams evaluating AI coding tools, that's worth more than feature comparisons. A tool that its own creators run on their own infrastructure tells you something about its reliability at the workflow level, not just the code-generation level.

The workload identity federation addition is also worth watching. Most teams running AI coding tools in CI pipelines have credential management problems. Workload identity federation removes the need to store static API keys, which is the kind of infrastructure detail that unlocks real enterprise deployment.

Example Use Case

A backend engineer working on a Python API needs to refactor a set of database query functions from raw SQL to SQLAlchemy ORM calls. They navigate to the project root, run claude, and describe the task in plain English: refactor the query functions in db/queries.py to use SQLAlchemy ORM instead of raw SQL strings, keeping return types the same.

Claude Code reads the existing file, generates the refactored code, applies the changes, and stages the diff. The engineer reviews it in the terminal, asks Claude Code to commit with a descriptive message, and moves on. The entire cycle stays inside the terminal with no browser tab involved.

This is the workflow I'd demo to a team still copy-pasting between chat windows and their editor. The round trip disappears.

Competitive Context

Most AI coding tools are IDE extensions or browser-based chat interfaces. Claude Code takes a different position: it runs where shell-centric developers already work, and treats the terminal as the primary surface rather than an afterthought.

The plugin directory and .claude/commands let teams define shared workflows that every developer's Claude Code session picks up automatically. Combined with the MDM deployment templates in the examples directory, that's a real path to standardizing AI-assisted workflows at the org level rather than the individual level.

The language breakdown in the repo is also worth noting: the codebase is now 79.7% Python, up from its earlier Shell-heavy composition. That shift toward Python suggests the tool's internal logic is moving toward more sophisticated scripting and automation capabilities.

My Take

131K stars and Anthropic's own team using it in CI are two different kinds of evidence, and both point in the same direction. Claude Code has moved past the "interesting experiment" phase.

If your team works in the terminal and wants AI assistance without leaving it, this is the most mature option I've seen for that workflow. The workload identity federation update is the one I'd watch most closely for enterprise teams still figuring out how to deploy AI coding tools safely in CI.