Agents & Models

Configure AI models, manage agent behavior, and customize your AI assistant.

AI Providers

Ollama (Local)

Run AI models locally on your machine. Private, free, and works offline.

ollama pull llama3.2
ollama pull qwen2.5

Cloud Providers

Connect to cloud AI services for more powerful models.

  • OpenAI (GPT-4, GPT-3.5)
  • Anthropic (Claude)
  • OpenRouter (Multiple models)

Model Configuration

~/.jarvis/openclaw.json:

{
  "agents": {
    "defaults": {
      "model": "ollama/llama3.2",
      "workspace": "~/.jarvis/workspace"
    }
  }
}

Configure your default model in the config file. Use jarvis configure for interactive setup.

Agent Behavior

Session Management

Jarvis maintains conversation context across channels. Each session stores your chat history and preferences.

Workspace

The workspace is where Jarvis stores files, sessions, and skills. Default location: ~/.jarvis/workspace

export JARVIS_CONFIG_DIR=~/.jarvis

Multi-Agent Routing

Run multiple agents with different configurations. Useful for separating personal and work contexts.

Popular Local Models

llama3.2 (4GB)

Meta's latest model, great balance of speed and quality

qwen2.5 (4GB)

Alibaba's model, excellent for coding tasks

llama3.2:1b (1GB)

Smaller, faster model for low-resource systems

mistral (4GB)

Strong general-purpose model

Jarvis

© 2026 Nilkanth Desai. MIT licensed.