AgentCrew Configuration Guide

AgentCrew Configuration Guide

This guide explains how to configure AgentCrew to work with different AI providers, customize agents, set up tools, and manage system settings.

Configuration Files

AgentCrew stores all configuration in `~/.AgentCrew/` (or `%USERPROFILE%\.AgentCrew\` on Windows):

~/.AgentCrew/
├── config.json              # Global settings and API keys
├── agents.toml              # Agent definitions and configurations
├── mcp_servers.json         # Model Context Protocol server configs
├── persistents/
│   └── adaptive.json        # Learned behaviors for all agents
└── conversations/           # Saved conversation history

Global Configuration (config.json)

The main configuration file controls API keys, UI preferences, and system behavior.

API Keys

Add your API keys to connect AgentCrew with AI providers:

{
  "api_keys": {
    "ANTHROPIC_API_KEY": "sk-ant-...",
    "OPENAI_API_KEY": "sk-proj-...",
    "GEMINI_API_KEY": "AIza...",
    "GITHUB_COPILOT_API_KEY": "",
    "DEEPINFRA_API_KEY": "",
    "TAVILY_API_KEY": "tvly-...",
    "VOYAGE_API_KEY": "pa-...",
    "ELEVENLABS_API_KEY": ""
  }
}

**Required Keys:**

undefined

**Optional Keys:**

undefined

**Getting API Keys:**

undefined

Custom LLM Providers

Add OpenAI-compatible providers like llama.cpp, Ollama, or LM Studio:

{
  "custom_llm_providers": [
    {
      "name": "llama.cpp",
      "type": "openai_compatible",
      "api_base_url": "http://localhost:8009",
      "api_key": "",
      "default_model_id": "qwen-14b",
      "available_models": [
        {
          "id": "../Qwen3-14B-Q6_K.gguf",
          "name": "Local Qwen 14B",
          "provider": "llama.cpp",
          "capabilities": ["thinking", "vision"],
          "default": true,
          "description": "Quantized Qwen model running locally",
          "input_token_price_1m": 0.0,
          "output_token_price_1m": 0.0
        }
      ],
      "is_stream": false,
      "extra_headers": {}
    }
  ]
}

**Fields:**

undefined