Skip to main content

Configuration

Exorous is highly customizable, allowing you to tailor its behavior to your specific project needs.

The config.toml File​

Exorous stores its configuration in ~/.exorous/config.toml (Global) and .ai-agent/config.toml (Project-specific). Project-level settings take precedence.

Core Settings​

# The model to use
model_name = "anthropic/claude-3-5-sonnet"

# Temperature for the LLM
temperature = 0.0

# How much context to keep before compression (in tokens)
max_context_tokens = 40000

# The default approval policy
approval_policy = "on-request"

Environment Variables​

You can also use environment variables to override settings or provide API keys:

  • EXOROUS_API_KEY: Your primary LLM provider key.
  • EXOROUS_MODEL: The model name.
  • EXOROUS_CWD: Set the default working directory.

Personalization​

Make Exorous work exactly how you like by tailoring its instructions and behavior.

  • Project Conventions: Set rules for indentation, file naming, or architectural patterns in your config.toml.
  • Personal Preferences: Define your favorite LLM models and interaction style.
  • Scoped Knowledge: Use the memory tool to store project-specific notes that the agent will automatically load at the start of every session.

Custom Instructions​

You can provide global or project-specific instructions to Exorous:

  • User Instructions: Tell Exorous about your preferences ("Always use tab for indentation", "Follow the Airbnb style guide").
  • Developer Instructions: Technical guidelines ("We use FastAPI for all microservices", "Database migrations must be handled via Alembic").

Tool Allowlisting​

Limit the tools Exorous has access to for specific projects:

[tools]
allowlist = ["read_file", "write_file", "grep", "code_search"]
# If allowlist is empty, all built-in tools are enabled.