{"_id":"@cyberboyayush/zenox-claude","name":"@cyberboyayush/zenox-claude","dist-tags":{"latest":"1.0.0"},"versions":{"1.0.0":{"name":"@cyberboyayush/zenox-claude","version":"1.0.0","description":"Claude Code plugin — Multi-agent orchestrator with Explorer (Haiku), Librarian (Sonnet), and Oracle (Opus) for intelligent task delegation, parallel research, and code review","author":{"name":"Ayush Sharma"},"license":"MIT","keywords":["claude-code","plugin","zenox","agents","orchestration","multi-agent","explorer","librarian","oracle","code-review"],"repository":{"type":"git","url":"git+https://github.com/CYBERBOYAYUSH/zenox-claude.git"},"publishConfig":{"access":"public"},"gitHead":"839e2e44d34bdf44f313828fdd5b98712b18255a","_id":"@cyberboyayush/zenox-claude@1.0.0","bugs":{"url":"https://github.com/CYBERBOYAYUSH/zenox-claude/issues"},"homepage":"https://github.com/CYBERBOYAYUSH/zenox-claude#readme","_nodeVersion":"22.12.0","_npmVersion":"11.12.0","dist":{"integrity":"sha512-1AYiZotIKT/xguUdu1aai5PoruEyEvjcz3hMngie7iPZejd+vnI8TxGseBBL5JG6fr2MkcJHxn6UWtqbhx1H3g==","shasum":"e8c3b03bbc26eb3a860df400c658fb7f4b54ebd6","tarball":"https://registry.npmjs.org/@cyberboyayush/zenox-claude/-/zenox-claude-1.0.0.tgz","fileCount":16,"unpackedSize":33777,"signatures":[{"keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U","sig":"MEUCIGeU5hSP6K4xnKr9/VyzQrNLUdxk/LgKzvdHL1LTiBRkAiEAvpP47LYH8COBzxtRP2A796LDHrYgDp7/sb1ADiBiJFs="}]},"_npmUser":{"name":"cyberboyayush","email":"aayush240204@gmail.com"},"directories":{},"maintainers":[{"name":"cyberboyayush","email":"aayush240204@gmail.com"}],"_npmOperationalInternal":{"host":"s3://npm-registry-packages-npm-production","tmp":"tmp/zenox-claude_1.0.0_1775987396449_0.983622076472318"},"_hasShrinkwrap":false}},"time":{"created":"2026-04-12T09:49:56.350Z","1.0.0":"2026-04-12T09:49:56.605Z","modified":"2026-04-12T09:49:56.821Z"},"maintainers":[{"name":"cyberboyayush","email":"aayush240204@gmail.com"}],"description":"Claude Code plugin — Multi-agent orchestrator with Explorer (Haiku), Librarian (Sonnet), and Oracle (Opus) for intelligent task delegation, parallel research, and code review","homepage":"https://github.com/CYBERBOYAYUSH/zenox-claude#readme","keywords":["claude-code","plugin","zenox","agents","orchestration","multi-agent","explorer","librarian","oracle","code-review"],"repository":{"type":"git","url":"git+https://github.com/CYBERBOYAYUSH/zenox-claude.git"},"author":{"name":"Ayush Sharma"},"bugs":{"url":"https://github.com/CYBERBOYAYUSH/zenox-claude/issues"},"license":"MIT","readme":"# Zenox for Claude Code\n\nMulti-agent orchestrator plugin for [Claude Code](https://docs.anthropic.com/en/docs/claude-code). Gives you a team of specialized AI agents instead of a single generalist.\n\n## What it does\n\nZenox replaces the default Claude Code agent with an orchestrator that automatically delegates to specialists:\n\n| Agent | Model | Role |\n|-------|-------|------|\n| **Zenox** (main) | Opus | Orchestrates everything, writes code, delegates when needed |\n| **Explorer** | Haiku | Fast codebase search — finds files, patterns, symbols |\n| **Librarian** | Sonnet | External research — official docs, GitHub examples, library internals |\n| **Oracle** | Opus | Strategic advisor — architecture decisions, debugging, code review |\n\nExplorer and Librarian run in **background** automatically. Oracle runs in **foreground** (blocks until result arrives). You don't manage the agents — the orchestrator delegates based on what you ask.\n\n## Install\n\n### From marketplace (recommended)\n\n```\n/plugin marketplace add CyberBoyAyush/zenox-claude\n/plugin install zenox@zenox-marketplace\n/reload-plugins\n```\n\n### From npm\n\nIf the marketplace is already configured to use npm:\n\n```\n/plugin marketplace add CyberBoyAyush/zenox-claude\n/plugin install zenox@zenox-marketplace\n```\n\n### Local development\n\n```bash\nclaude --plugin-dir /path/to/zenox-claude\n```\n\n## How delegation works\n\nYou don't call agents directly. The main agent delegates automatically:\n\n```\nYou: \"Where's the authentication logic?\"\n-> Explorer searches the codebase (background)\n\nYou: \"How does React Query handle caching?\"\n-> Librarian fetches docs + real GitHub examples (background)\n\nYou: \"Should I use Redux or Zustand here?\"\n-> Oracle analyzes trade-offs (foreground)\n\nYou: \"review\"\n-> Oracle reviews your recent git diff for bugs, security, regressions\n```\n\n## Skills (keyword triggers)\n\n| Skill | Trigger | What happens |\n|-------|---------|--------------|\n| `/ultrawork` | \"ultrawork\" or \"ulw\" | Maximum multi-agent coordination — fires parallel agents |\n| `/deep-research` | \"deep research\" | 3-4 agents explore internal + external sources |\n| `/explore-codebase` | \"explore codebase\" | Multiple explorers map out relevant code |\n| `/review` | \"review\", \"self-review\" | Oracle reviews your git diff |\n\n## Included MCP servers\n\nThese are auto-loaded when the plugin is enabled:\n\n| Server | Purpose |\n|--------|---------|\n| **exa** | Web search, docs lookup, URL crawling (free, no API key needed) |\n| **grep-app** | Search millions of public GitHub repos |\n| **sequential-thinking** | Step-by-step reasoning for complex problems |\n\nAll three servers are publicly accessible and require no API keys or configuration.\n\n## Self-review protocol\n\nAfter significant implementations (3+ files changed, security-sensitive code, architecture changes), the orchestrator automatically:\n\n1. Runs `git diff` to capture changes\n2. Fires Oracle with the complete diff\n3. Collects the review verdict\n4. Addresses Critical/High findings before delivering\n\nSkipped for trivial changes (single file, <50 lines, docs/config only).\n\n## Plugin structure\n\n```\nzenox-claude/\n├── .claude-plugin/\n│   ├── plugin.json          # Plugin manifest\n│   └── marketplace.json     # Marketplace catalog\n├── agents/\n│   ├── zenox.md             # Main orchestrator agent\n│   ├── explorer.md          # Codebase search specialist\n│   ├── librarian.md         # External research specialist\n│   └── oracle.md            # Strategic advisor + code reviewer\n├── skills/\n│   ├── ultrawork/SKILL.md\n│   ├── deep-research/SKILL.md\n│   ├── explore-codebase/SKILL.md\n│   └── review/SKILL.md\n├── hooks/\n│   └── hooks.json\n├── .mcp.json                # MCP server configs (exa, grep-app, sequential-thinking)\n├── settings.json            # Activates zenox as main agent\n└── package.json             # npm distribution\n```\n\n## Configuration\n\nThe plugin activates the `zenox` agent as the main thread via `settings.json`. Model assignments are fixed to Claude's available tiers:\n\n- **Explorer**: Haiku (fast, cheap — right for grep-like tasks)\n- **Librarian**: Sonnet (balanced — good for research synthesis)\n- **Oracle**: Opus (maximum reasoning — right for architecture/review)\n- **Zenox**: Opus (main thread needs full capability)\n\nClaude Code only supports Anthropic models (haiku/sonnet/opus) for agents, so there's no model configuration step.\n\n## Differences from Zenox for OpenCode\n\nThis is a port of [Zenox](https://github.com/CyberBoyAyush/zenox) (OpenCode plugin) adapted for Claude Code's declarative plugin system:\n\n| Aspect | OpenCode Zenox | Claude Code Zenox |\n|--------|---------------|-------------------|\n| Implementation | TypeScript plugin (~1000 LOC) | Static markdown files |\n| Agent count | 4 (includes UI Planner) | 3 (no UI Planner) |\n| Model flexibility | Any provider (OpenAI, Google, Anthropic) | Anthropic only |\n| Background tasks | Custom BackgroundManager | Native `background: true` frontmatter |\n| Model config | Interactive CLI (`bunx zenox config`) | Fixed in agent frontmatter |\n| Distribution | npm + OpenCode plugin registry | npm + Claude Code marketplace |\n\n## License\n\n[MIT](LICENSE)\n","readmeFilename":"README.md","_rev":"1-c429f67f687d533557e786fe567eaf65"}