tachibot-mcp

Stop AI Hallucinations Before They Start Run models from OpenAI, Google, Anthropic, xAI, Perplexity, and OpenRouter in parallel. They check each other's work, debate solutions, and catch errors before you see them.

GitHub

TachiBot MCP - Universal AI Orchestrator

Version License Node MCP

Multi-model AI orchestration platform with 31 tools, advanced workflows, and intelligent prompt engineering. Works with Claude Code, Claude Desktop, Cursor, and any MCP-compatible client.

🔑 Bring Your Own Keys

Two ways to connect — pick what works for you:

ModeKeys NeededBest For
Gateway Mode2 keys (OpenRouter + Perplexity)Simplicity, unified billing
Direct Mode5+ keys (one per provider)Maximum control, no middleman

Gateway Mode (Recommended)

Route GPT-5, Gemini, and Grok through OpenRouter — same key you need for Qwen/Kimi anyway.

OPENROUTER_API_KEY=sk-or-xxx     # GPT-5, Gemini, Grok, Qwen, Kimi
PERPLEXITY_API_KEY=pplx-xxx      # Required separately (not on OpenRouter)
USE_OPENROUTER_GATEWAY=true

Direct Mode

Use each provider's native API directly.

OPENAI_API_KEY=sk-xxx            # GPT-5
GOOGLE_API_KEY=xxx               # Gemini
GROK_API_KEY=xxx                 # Grok
OPENROUTER_API_KEY=sk-or-xxx     # Qwen, Kimi (always via OpenRouter)
PERPLEXITY_API_KEY=pplx-xxx      # Perplexity

Both modes work. Gateway mode just means fewer accounts to manage.

Get keys: OpenRouter | Perplexity

📚 Documentation

🌐 Full Documentation: tachibot.com/docs

Quick Links

✨ Key Features

🤖 Multi-Model Intelligence

  • 31 AI Tools: Perplexity, Grok, GPT-5, Gemini, Qwen, Kimi (32 with competitive mode)
  • Multi-Model Reasoning: Challenger, Verifier, Scout modes
  • Smart Routing: Automatic model selection for optimal results
  • OpenRouter Gateway: Optional single API key for all providers (OpenAI, Gemini, Grok)

🔄 Advanced Workflows

  • YAML-Based Workflows: Define complex multi-step AI processes
  • Prompt Engineering: 14 research-backed techniques built-in
  • Auto-Synthesis: Prevents token overflow on large workflows
  • Parallel Execution: Run multiple models simultaneously

🎯 Tool Profiles

  • Minimal (8 tools) - Budget-friendly, token-constrained
  • Research Power (13 tools) - Default, best balance
  • Code Focus (15 tools) - Software development
  • Balanced (18 tools) - General-purpose
  • Full (31 tools, includes competitive mode) - Maximum capability

🔧 Developer Experience

  • Claude Code Native - First-class support
  • Claude Desktop - Full integration
  • Cursor - Works seamlessly
  • TypeScript - Fully typed
  • Extensible - Add custom tools & workflows

🚀 Quick Start

Installation

# Via NPM (recommended)
npm install -g tachibot-mcp

# Verify installation
tachibot --version

Setup

Option A: Single Key Mode (Recommended)

{
  "mcpServers": {
    "tachibot": {
      "command": "tachibot",
      "env": {
        "OPENROUTER_API_KEY": "sk-or-xxx",
        "PERPLEXITY_API_KEY": "pplx-xxx",
        "USE_OPENROUTER_GATEWAY": "true"
      }
    }
  }
}

Option B: Direct Provider Keys

{
  "mcpServers": {
    "tachibot": {
      "command": "tachibot",
      "env": {
        "PERPLEXITY_API_KEY": "your-key",
        "GROK_API_KEY": "your-key",
        "OPENAI_API_KEY": "your-key",
        "GOOGLE_API_KEY": "your-key",
        "OPENROUTER_API_KEY": "your-key"
      }
    }
  }
}

Restart Claude and you're ready!

See Installation Guide for detailed instructions.

📦 What's Included

Core Tools

  • 🔍 Research: perplexity_ask, perplexity_research, grok_search, scout
  • 🧠 Reasoning: grok_reason, openai_reason, perplexity_reason, kimi_thinking, focus
  • 💡 Analysis: gemini_brainstorm, gemini_analyze_text, openai_brainstorm, qwen_coder
  • ✅ Validation: verifier, challenger (*multi-model workflows)
  • 💭 Meta: think, nextThought

Advanced Modes

  • Focus - Deep collaborative reasoning (4+ models)
  • Scout - Multi-source information gathering
  • Challenger - Critical analysis with fact-checking
  • Verifier - Multi-model consensus verification

Workflows

  • Ultra Creative Brainstorm - 15 steps, 10 prompt techniques
  • Code Architecture Review - Systematic code analysis
  • Code Review - PR and code review workflow
  • UX Research Flow - User experience research
  • Accessibility Audit - WCAG compliance checking
  • Creative Brainstorm - Quick ideation workflow
  • PingPong - Multi-model debate/collaboration
  • System Workflows - verifier, challenger, scout
  • Custom Workflows - Build your own in YAML

🌟 Example Usage

Quick Research

// In Claude Code
perplexity_ask({
  query: "What are the latest developments in transformer architecture?"
})

Multi-Model Reasoning

focus({
  query: "Design a scalable microservice architecture",
  mode: "deep-reasoning",
  models: ["grok-3", "gpt-5", "gemini-2.5"],
  rounds: 5
})

Run Workflow

workflow --name ultra-creative-brainstorm --query "AI-powered code review system"

🎓 Learn More

Documentation

Setup Guides

🤝 Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

📄 License

AGPL-3.0 - see LICENSE for details.

🔗 Links

Made with ❤️ by @byPawel

Transform your AI workflow with intelligent multi-model orchestration.

Related Servers