The AI Development Platform
_{Governance. Agents. Dashboard. Terminal. All in one.}

21 commands · 22 agents · 58 UI components · Infinity Terminal · MCP dashboard
Full-stack AI development platform built on Claude Code

---

What is Forge?

Forge is a full-stack AI development platform that turns Claude Code into a Chief of Staff for your codebase. It combines three systems:

1. Plugin — 21 slash commands, 22 specialized agents, 29 skills, and governance hooks that extend Claude Code natively
2. Dashboard — React-based web UI with real-time governance HUD, agent activity feed, and project health monitoring
3. Infinity Terminal — Browser-based terminal with PTY bridge and session persistence that survives disconnects

No lock-in. The plugin works standalone without the dashboard. The dashboard adds visual oversight for teams and power users.

Quick Start

Plugin Only (Zero Dependencies)

# Install from Claude Code
claude plugin add nxtg-ai/forge-plugin

# Initialize in your project
/[FRG]-init

# Check health
/[FRG]-status

Full Platform (Dashboard + Terminal)

# Clone the repo
git clone https://github.com/nxtg-ai/forge.git
cd forge

# Install dependencies
npm install

# Start everything
npm run dev
# Dashboard:  http://localhost:5050
# API:        http://localhost:5051
# Terminal:   http://localhost:5050/terminal

The Platform

Governance Dashboard

Real-time project health monitoring with a dark-themed, responsive UI:

• Health Score — A/B/C/D/F grade computed from 8 quality checks
• Governance HUD — Live project context, workstream tracking, blocker detection
• Agent Activity Feed — Watch 22 agents coordinate in real-time
• Oracle Feed — Governance events and strategic insights
• Command Center — Execute and monitor Forge commands from the browser
• Vision Display — Project vision, goals, and alignment tracking

58 React components. Framer Motion animations. Tailwind CSS. Fully responsive.

Infinity Terminal

A browser-based terminal that never dies:

• Sessions persist through browser close, network disconnects, and server restarts
• Multiple clients connect to the same session simultaneously
• Full PTY bridge with xterm.js 6.0
• Works over Vite proxy for multi-device access (phones, tablets, remote machines)

Browser (xterm.js) → WebSocket → PTY Bridge → Shell
                          ↓
               Session persistence via:
               - Session ID tracking
               - Auto-reconnect
               - State serialization

21 Slash Commands

Governance
/[FRG]-init	60-second setup wizard
/[FRG]-status	Git, tests, types, governance at a glance
/[FRG]-status-enhanced	Full dashboard with health score and trends
/[FRG]-gap-analysis	Test, doc, security, and architecture gaps
/[FRG]-compliance	License scanning and SBOM generation
/[FRG]-command-center	Central hub for all capabilities
/[FRG]-dashboard	Open visual governance dashboard (MCP-powered)
Feature Development
/[FRG]-feature	Plan and build with full agent orchestration
/[FRG]-spec	Generate detailed technical specifications
/[FRG]-agent-assign	Assign tasks to the 22-agent roster
/[FRG]-integrate	Scaffold third-party integrations
Quality
/[FRG]-test	Run tests with failure analysis
/[FRG]-deploy	Pre-flight validation before shipping
/[FRG]-optimize	Performance and maintainability analysis
/[FRG]-upgrade	Config gap detection and drift repair
State & Docs
/[FRG]-checkpoint	Save a named project snapshot
/[FRG]-restore	Roll back to any checkpoint
/[FRG]-report	Session activity and progress reports
/[FRG]-docs-status	Documentation health and coverage
/[FRG]-docs-update	Detect and fix stale documentation
/[FRG]-docs-audit	Full documentation quality audit

22 Specialized Agents

Agent	Specialty	Agent	Specialty
CEO-LOOP	Autonomous strategic decisions	Oracle	Real-time governance sentinel
Orchestrator	Multi-agent coordination	Planner	Architecture and task breakdown
Builder	Feature implementation	Guardian	Quality gates and validation
Detective	Root cause analysis	Security	Vulnerability scanning
Testing	Test generation and coverage	Compliance	License and regulatory checks
Docs	Documentation generation	Refactor	Code restructuring
Performance	Profiling and optimization	Analytics	Metrics and reporting
Learning	Pattern recognition	UI	Frontend development
API	Endpoint design	Database	Schema and queries
DevOps	CI/CD and infrastructure	Integration	External services
Release Sentinel	Doc sync on release	Gov. Verifier	Governance validation

5 Engagement Modes

Mode	You see	Best for
CEO	Results only. Agents decide and execute autonomously.	Shipping fast
VP	Decisions surfaced. Details on demand.	Strategic oversight
Engineer	Full agent activity and technical logs.	Hands-on development
Builder	Implementation details and task queues.	Deep coding sessions
Founder	Everything. No filters. Full transparency.	Debugging and learning

MCP Governance Tools

8 tools available via Model Context Protocol:

Tool	What it does
forge_get_health	Health score with letter grade
forge_get_governance_state	Read governance.json
forge_get_git_status	Branch, commits, working tree
forge_get_code_metrics	Files, tests, coverage, dependencies
forge_run_tests	Auto-detect and run test suite
forge_list_checkpoints	List saved snapshots
forge_security_scan	Secrets, eval, .env, npm audit
forge_open_dashboard	Generate + open HTML dashboard

Architecture

NXTG-Forge
  ├── src/
  │   ├── components/          # 58 React components (17K lines)
  │   │   ├── governance/      #   Governance HUD, Oracle, Agent Activity
  │   │   ├── dashboard/       #   Health scores, metrics, charts
  │   │   ├── infinity-terminal/#  xterm.js terminal with PTY bridge
  │   │   └── command/         #   Command center, output drawer
  │   ├── server/              # Express API + WebSocket server
  │   │   ├── api-server.ts    #   REST API, rate limiting, CORS
  │   │   └── workers/         #   Agent worker pool infrastructure
  │   ├── adapters/            # Multi-backend AI CLI adapters
  │   │   ├── claude-code.ts   #   Claude Code adapter
  │   │   └── factory.ts       #   Auto-detection factory
  │   ├── services/            # Business logic layer
  │   ├── hooks/               # 13 custom React hooks
  │   ├── pages/               # 25 page components
  │   └── contexts/            # React context providers
  │
  ├── .claude/
  │   ├── commands/            # 21 slash commands (markdown)
  │   ├── agents/              # 22 agent definitions
  │   ├── skills/              # Knowledge modules
  │   └── governance.json      # Project state
  │
  └── forge-plugin/            # Standalone Claude Code plugin
      └── (see github.com/nxtg-ai/forge-plugin)

Multi-Backend Support

Backend	Status
Claude Code	Full support — commands, agents, skills, hooks, dashboard
Codex CLI	Framework skill + adapter (in progress)
Gemini CLI	Framework skill + adapter (in progress)

Development

# Install
npm install

# Start dev server (dashboard + API)
npm run dev

# Run tests
npm test

# Type check
npx tsc --noEmit

# Quality gates (tests + types + lint)
npm run quality:gates

Test Coverage

Metric	Value
Test files	114
Tests passing	4,085
TypeScript errors	0 (strict mode)
Source files	235
Total lines	137,204

Technology Stack

All dependencies MIT or Apache-2.0 licensed.

Layer	Technology
Frontend	React 19, Vite 7, Tailwind CSS, Framer Motion
Backend	Express 4, WebSocket, node-pty
Terminal	xterm.js 6.0 with PTY bridge
Testing	Vitest, Testing Library
Plugin	Pure markdown (zero dependencies)
MCP	@modelcontextprotocol/sdk

Related

• forge-plugin — Standalone Claude Code plugin (commands + agents + skills only, no dashboard)

Contributing

1. Fork and clone
2. npm install && npm test
3. Make changes with tests
4. npm run quality:gates
5. Open a PR

License

MIT — Copyright (c) 2026 NXTG AI

---

_{Built for developers who ship. Governed by Forge.}