Quick Start (L1 Vibe Coder)
Get AI-powered governance inside Claude Code in 30 seconds. No dependencies. No config. No runtime.
Get AI-powered governance inside Claude Code in 30 seconds. No dependencies. No config. No runtime.
Install
claude plugin marketplace add nxtg-ai/forge-plugin
claude plugin install forge
That's it. The plugin loads 22 agents, 21 slash commands, 29 knowledge skills, and 6 governance hooks. Everything is pure markdown — no build step, no compilation.
Your First Command
Open Claude Code in any project and run:
/forge:status
You'll see:
- Project health score — A through F grade based on 8 quality dimensions
- Git status — branch, uncommitted changes, recent commits
- Test results — test count, pass rate, coverage (if available)
- Security scan — patterns for secrets, unsafe code, dependency vulnerabilities
Find What's Missing
/forge:gap-analysis
This scans your project across 5 dimensions and tells you exactly what's missing:
| Dimension | What It Checks |
|---|---|
| Testing | File ratio, coverage reports, missing test files |
| Documentation | README presence, inline docs, API documentation |
| Security | Hardcoded secrets, eval/exec usage, npm audit |
| Architecture | File organization, large files, circular patterns |
| Performance | Bundle size, dependency count, build configuration |
Each gap comes with an actionable recommendation and priority level.
Build a Feature
/forge:feature "add user authentication"
This spawns a multi-agent team inside Claude Code:
- Planner designs the architecture and breaks work into steps
- Builder implements the feature with tests
- Guardian runs quality checks on the output
- Security reviews for vulnerabilities
You watch. They coordinate. The output is production-ready code with tests.
What's Running Behind the Scenes
The plugin installs an MCP governance server that provides 8 tools:
| Tool | What It Does |
|---|---|
forge_get_health | Composite health score (0-100, A-F grade) |
forge_get_git_status | Branch, commits, clean status |
forge_get_code_metrics | Lines, files, dependencies, test ratio |
forge_run_tests | Auto-detects runner (vitest/jest/pytest) |
forge_security_scan | Pattern-based secret and vulnerability detection |
forge_list_checkpoints | Saved governance snapshots |
forge_get_governance_state | Full governance state object |
forge_open_dashboard | Generate and launch HTML health dashboard |
These tools run automatically — agents and hooks call them as needed. You don't invoke them directly.
Key Commands
| Command | What It Does |
|---|---|
/forge:status | Project health at a glance |
/forge:gap-analysis | Find missing tests, docs, security gaps |
/forge:feature "desc" | Multi-agent feature development |
/forge:test | Run tests with analysis |
/forge:checkpoint | Save restorable project state |
/forge:deploy | Deploy with pre-flight checks |
/forge:docs-audit | Check documentation coverage |
When to Upgrade
L1 works standalone. You never need to install anything else.
But if you start running multiple AI tools on the same codebase (Claude Code + Codex + Gemini), you'll hit file conflicts and lost context. That's when L2 Pro Builder earns its keep.
Signs you need L2:
- Two AI tools edited the same file and one overwrote the other
- You're manually copy-pasting context between tools
- You want a task board that tracks what each tool is working on
- You need decisions and patterns to persist across sessions
Next Steps
- Upgrade to Pro Builder (L2) — Multi-tool orchestration
- Slash Commands Reference — All 21 commands in detail
- Agents Reference — When to use each of the 22 agents