Ori - OpenRouter Agent Tools

Ori is OpenRouter’s family of tools for putting agents to work with any model. One CLI, ori, signs you in with your OpenRouter account and gives every tool access to the models, credits, and org settings you already have.

The Ori tools

Tool	What it does	Status
Ori Harness	Runs the agent CLI you already use, such as Claude Code, Codex, OpenCode, or Hermes, on OpenRouter with one command. See the Ori Harness guide.	Available
Ori Code	OpenRouter’s native terminal coding agent, built for working with many models.	Available
Ori Eval	Finds the best model for your project by running your agent on your own prompts and grading the results. See the Ori Eval guide.	Available
Ori Intern	Slack agents that learn skills and run on any model. See the Intern chat guide.	Waitlist
Ori Desktop	Send tasks to any OpenRouter model from the Mac menu bar.	Waitlist

Join the waitlist for Ori Intern and Ori Desktop at openrouter.ai/ori.

Install Ori

Install the CLI, then sign in with your OpenRouter account. For interactive use you don’t need an API key.

macOS, Linux, WSL
Windows PowerShell

curl -fsSL https://openrouter.ai/labs/ori/install.sh | bash

irm https://openrouter.ai/labs/ori/install.ps1 | iex

ori login

Then run the tool you want:

ori claude          # Claude Code on OpenRouter (or codex, opencode, hermes, ...)
ori code -p "What is the best model for my support agent?"   # writes and runs an eval, then recommends a model

After the agent writes the eval, ori eval runs every *.eval.ts file in your project again, locally or in CI.

Next steps

Ori Harness: run your agent CLI on OpenRouter.
Ori Eval: write and run evals that pick the best model.
Configuration: settings, models, and defaults for every Ori tool.
Changelog: what changed in each Ori release.