AI Coding Agents · sangrokjung

claude-forge

Claude Forge is a plugin framework that extends Claude Code (Anthropic's AI coding assistant) with 11 specialist agents, 34 slash commands, 25 reusable skills, and 15 safety hooks. It's positioned as a productivity layer similar to oh-my-zsh for terminal shells—pre-configured workflows for planning, testing, code review, and security checks.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/sangrokjung/claude-forge

774

GitHub stars

162

Forks

Shell

Primary language

MIT

License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	sangrokjung/claude-forge
Owner	sangrokjung
Primary language	Shell
License	MIT — OSI-approved
Stars	774
Forks	162
Open issues	29
Latest release	v3.1.0 (2026-06-30)
Last updated	2026-06-30
Source	https://github.com/sangrokjung/claude-forge

What claude-forge is

A shell-based plugin framework for Claude Code that provides multi-agent orchestration (planner, architect, security-reviewer, tdd-guide, etc.), CLI command shortcuts, lifecycle hooks covering 21+ events, MCP connections (Playwright, jina-reader, context7, Chrome DevTools), and rule-based behavioral guidelines. Installable via plugin marketplace or direct bash/PowerShell script.

Quickstart

Get the claude-forge source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/sangrokjung/claude-forge.git2cd claude-forge3# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Accelerating full-stack development workflows

Teams using Claude Code benefit from automated plan→test→review→verify pipeline, reducing manual context-switching and ensuring consistent TDD, security, and code-review practices across sessions.

Enforcing security and compliance guardrails

The 6-layer hook system and security-reviewer agent catch OWASP vulnerabilities, secret leaks, and unsafe database commands before they reach production, suitable for regulated or security-sensitive projects.

Standardizing coding practices across teams

Pre-built agents (database-reviewer, doc-updater, refactor-cleaner) and 9 rule files embed company conventions and coding standards, so Claude Code behaves consistently with team expectations without per-session prompting.

Implementation considerations

Option A (plugin marketplace) provides partial features (commands + some skills); Option B (direct install) required for full agent, hook, rule, and MCP access. Team must decide upfront whether lightweight or comprehensive setup is needed.
Installation via curl | bash (Option B) downloads and executes code directly. Security review of install.sh and git submodules should precede production deployment, especially in regulated environments.
MCP connections (Playwright, jina-reader, Chrome DevTools) require external tool availability and configuration. Browser automation and live docs search are not enabled by default and need setup.
Hooks system runs silently on 21+ lifecycle events. Misconfigured or overly restrictive hooks could block legitimate workflows; test hook rules in staging before rollout.
Rules (9 behavior files) are read at session start. Updates require new Claude Code session; existing conversations do not reload rules—team must establish session refresh practices.

When to avoid it — and what to weigh

You need Claude Code integration with non-Anthropic LLMs — Claude Forge is tightly coupled to Claude Code's API and agent protocol. It does not adapt to GPT, Gemini, or other proprietary AI platforms.
Your team uses Windows-only environments without PowerShell Admin access — Full install requires bash (Option B) or PowerShell as Administrator (Windows). Option A (plugin marketplace) has partial feature coverage and may be limiting.
You need a stable, long-term LTS release cycle — Latest release (v3.1.0, June 2026) references recent Anthropic API changes and 'Hooks 21+ events'. Requires active monitoring of Claude Code API stability and backward compatibility across upgrades.
Your workflows don't fit the predefined agent/skill model — Claude Forge provides fixed agents (planner, architect, security-reviewer) and 25 canned skills. Custom agents or highly specialized workflows require forking and modifying the framework itself.

License & commercial use

MIT License. Permissive open-source; allows commercial use, modification, distribution, and private use. Standard disclaimer: provided as-is, no warranty.

MIT is a standard permissive OSI license that explicitly permits commercial use. No license restrictions prevent commercial deployment. However, Claude Forge depends on Claude Code (Anthropic), which has its own terms of service and commercial licensing; ensure your Claude Code usage aligns with Anthropic's commercial policy.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Strong
License clarity	Clear
Deployment complexity	Low
DEV.co fit	Strong
Assessment confidence	High

Security considerations

Security hooks cover 15 built-in + 9 optional checks including password leak detection, database safety, and remote script validation. However, 'security' claims in project description are not independently verified. Hooks operate at policy level (e.g., regex patterns, API calls); their effectiveness depends on correct configuration and Claude Code's ability to follow hook directives. Install via curl | bash downloads executable code; use requires trust in upstream repo and GitHub delivery. No mention of code signing, audit logs, or security audit history. Recommend: review install.sh before execution, run in staging, and validate hook behavior against your threat model.

Alternatives to consider

Continue.dev (open-source VS Code extension)

IDE-native AI coding assistant with plugin support and customizable workflows, but not Claude-specific and requires manual setup; less opinionated than Claude Forge.

Codeium (VS Code, JetBrains, Neovim)

General-purpose AI code completion with IDE integrations, but not tied to Claude Code; lacks agent-based orchestration and security hooks.

Anthropic's own Claude Code + custom prompting

Claude Code with hand-crafted system prompts and user-level workflows; free and fully customizable, but requires manual assembly and maintenance of agent logic, commands, and rules.

Software development agency

Build on claude-forge with DEV.co software developers

Claude Forge brings structured workflows, security guardrails, and specialist agents to Claude Code in under 5 minutes. If your team uses Claude Code and wants consistent, automated TDD and security practices, explore a full install with Option B or start lightweight with the plugin marketplace.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

claude-code-plugins-plus-skills claude-plugins-official claude-code-prompts

Related on DEV.co

Explore the category and the services that help you build with it.

AI Coding Agents AI development Web app development API development Talk to DEV.co

claude-forge FAQ

Can I use Claude Forge with GPT-4 or other LLMs?

No. Claude Forge is built specifically for Claude Code (Anthropic). The agent protocol, slash commands, and hooks are tightly coupled to Claude Code's API and will not transfer to other LLM providers.

What's the difference between Option A (plugin install) and Option B (direct install)?

Option A installs commands and partial skills via Claude Code's plugin marketplace—quick, cloud-hosted, limited. Option B runs a local bash/PowerShell script to install all 11 agents, 15 hooks, 9 rules, and 4 MCP connections—complete but requires terminal access and more setup.

Do I need to modify hook rules if I already have CI/CD or linting?

Not necessarily. Hooks are safety checks in Claude Code's runtime; CI/CD runs after code is committed. They can coexist. However, if your CI/CD already enforces rules (e.g., security scans, test coverage), overlapping hooks may be redundant. Review and tailor to avoid conflicts.

Is Claude Forge maintained and production-ready?

Active as of June 2026 (v3.1.0). 774 stars and ongoing commits suggest adoption. However, it depends on Claude Code API stability; Anthropic breaking changes or major version jumps could require framework updates. Suitable for production if your team can handle minor version upgrades.

Software development & web development with DEV.co

DEV.co is a software development agency delivering custom software development services to companies building on open source. Our software developers and web developers design, integrate, and ship production systems — spanning web development, APIs, AI, data, and cloud. If claude-forge is part of your ai coding agents roadmap, our team can implement, customize, migrate, and maintain it.

Ready to Streamline Your AI-Assisted Development?

Talk to DEV.co Schedule a Consultation