Setup

Configuration

Dojo works with zero configuration using sensible defaults. To customize behavior, create a dojo.toml in your project root:

[skills]
dir = ["skills", ".agents/skills"]

[model]
provider = "copilot"
evaluator = "gpt-4o"

[reporting]
per-skill = true
consolidated = false

See the Config Reference for the full schema.

Global CLI flags

These flags apply to all commands and override dojo.toml values:

Flag	Description
--cwd <dir>	Working directory for config and skill discovery
--skills-dir <dir>	Override skills directory (repeatable)
--model-provider <provider>	Override model provider
--evaluator-model <model>	Override evaluator model

See the Commands Reference for all available commands and options.

Model provider

Dojo uses a model provider to power the evaluator agent that runs your evals. Currently, GitHub Copilot is the only supported provider. It is used by default and requires no additional configuration beyond having Copilot access.

[model]
provider = "copilot"
evaluator = "gpt-4o"

The provider abstraction is designed to support additional providers in the future. See the Config Reference for the model configuration options.