Skip to main content
1

Install the CLI

Install the RELAI CLI with the standalone installer for macOS.
curl -fsSL https://api.relai.ai/install.sh | bash
2

Configure credentials

Run setup after installing. It checks prerequisites, asks for your RELAI API credentials, and saves your theme in ~/.relai/config.toml. It also installs plugins for any detected coding agents (Codex or Claude Code).
The default API URL is https://api.relai.ai.
Create or manage CLI keys from the API keys page. Never paste or commit API keys.
relai setup
3

Initialize your project

From your git-tracked agent directory, register the project and generate the simulator harness.
relai init
relai init registers the project and generates a custom simulator harness for your agent. See the project requirements for what RELAI needs from your project.

​
Prerequisites

RELAI expects these tools on your PATH. It creates a project-local simulator virtualenv at .relai/simulator/.venv that must run Python 3.11+.
ToolRequiredNotes
gitYesProject must be a git repo with ≥1 commit. Install
python3YesPython 3.11 or newer. Install
jqYesJSON processing. Install
uvRecommendedIsolated simulator env without touching system Python. Install
ghRecommendedEnables automatic optimizer PRs. Install

​
Project requirements

Before running relai init, make sure that:
  • the project is a git repository with at least one commit.
  • the project is structured as a valid Python module from which your agent can be imported.
  • the simulator environment can run on Python 3.11 or newer.
RELAI strives to be framework-agnostic, so your agent doesn’t have to use a specific library. Support is deepest today for OpenAI Agents SDK, LangChain, and LangGraph; other frameworks work on a best-effort basis that’s still improving.
RELAI is in beta. For where framework support is deepest, plus constraints and known limitations, see Compatibility.