Claude Forge — Plugin Framework for Claude Code

What it is

Claude Forge is a plugin framework for Claude Code, inspired by the oh-my-zsh model of community-driven extensions. It provides 11 AI agents, 36 commands, and 15 skills that extend Claude Code's capabilities. The framework includes 6-layer security hooks to validate agent actions before execution.

Claude Forge targets developers who want to customize and extend Claude Code with reusable plugins, shared workflows, and team-specific agents without modifying the core Claude Code installation.

How it saves time or tokens

Claude Forge pre-packages common development workflows as plugins. Instead of writing custom CLAUDE.md instructions for every project, you install a plugin that encapsulates the workflow with proper tool definitions, prompts, and guardrails.

The security hooks reduce token waste by catching invalid actions early in the pipeline before they reach the LLM for processing.

Additionally, the project's well-structured documentation and active community mean developers spend less time troubleshooting integration issues. When AI coding assistants generate code for this tool, they can reference established patterns from the documentation, producing correct implementations with fewer iterations and lower token costs.

How to use

1. Install Claude Forge:

git clone https://github.com/claude-forge/claude-forge.git ~/.claude-forge
cd ~/.claude-forge && ./install.sh

1. Enable plugins in your project:

claude-forge enable code-review
claude-forge enable git-workflow
claude-forge enable test-generator

1. Use the new commands in your Claude Code session:

/forge:review     -- Run code review with the code-review agent
/forge:test       -- Generate tests for the current file
/forge:deploy     -- Execute the deployment workflow

1. Create custom plugins by defining agents, commands, and hooks in a plugin directory.

Example

# Plugin definition: .claude-forge/plugins/my-team/plugin.yaml
name: my-team-workflow
agents:
  - name: reviewer
    prompt: 'Review code for our team standards'
    tools: [read, grep, glob]
commands:
  - name: review
    agent: reviewer
    description: 'Run team code review'
hooks:
  pre-commit:
    - validate-no-secrets
    - run-linter

Related on TokRepo

• AI Tools for Coding — AI coding assistants and development frameworks
• Featured Workflows — Discover more curated developer tools

Common pitfalls

• Installing too many plugins at once. Each plugin adds context to Claude Code sessions, consuming tokens. Start with 2-3 essential plugins and add more as needed.
• Not configuring security hooks for sensitive projects. The 6-layer security system is powerful but needs to be configured for your specific threat model.
• Treating plugins as static. Update plugins regularly as Claude Code capabilities evolve. Outdated plugins may use deprecated patterns.
• Failing to review community discussions and changelogs before upgrading. Breaking changes in major versions can disrupt existing workflows. Pin versions in production and test upgrades in staging first.