Loading documentation...
Control which tools are loaded to manage token usage. Each tool adds ~400-500 tokens to your context. Enable only what you need.
TachiBot provides 64 tools, but you rarely need all of them at once. By enabling only the tools you need, you can control token usage and costs:
think tool = ~400 tokensthink + focus + nextThought = ~1,200 tokensCreate or edit tools.config.json in your project root:
think - Structured reasoning scratchpadfocus - Multi-model orchestrationnextThought - Sequential thinkingjury - Multi-model jury panel with Gemini judge synthesisplan_critique - Adversarial red-team of an existing plan (pre-mortem, ranked risks)security_review - OWASP/CWE security audit with taint/data-flow analysisworkflow - Execute YAML/JSON workflows (e.g. --name pingpong-debate-3rounds for multi-model debates)list_workflows - List available workflowscreate_workflow - Create custom workflowsperplexity_ask - Web search with up-to-date informationperplexity_reason - Complex reasoning grounded in live web searchgrok_reason - First principles reasoninggrok_code - Code analysis and optimizationgrok_debug - Debugging assistancegrok_architect - System architecture and designgrok_brainstorm - Creative brainstorminggrok_search - Live web searchopenai_reason - Mathematical and scientific reasoning with GPT-5.5openai_brainstorm - Creative ideation using GPT-5.5gemini_brainstorm - Collaborative brainstorminggemini_analyze_code - Code quality and security analysisgemini_analyze_text - Text analysis and sentimentqwen_coder - Advanced code generation with Qwen3-Coder-Next (80B/3B MoE)See API Keys Setup Guide for detailed instructions on obtaining and configuring these API keys.
You can modify tools.config.json at any time. Restart TachiBot to apply changes:
think / focus / nextThought enabled for best performanceThe think tool alone provides 54% improvement in complex reasoning tasks (per Anthropic research). Always keep it enabled.