Public guide / Setup

Getting started

Connect a provider first, then give the agent a task it can inspect and verify.

1. Install your surface

Use the official VS Code Marketplace, Chrome Web Store, or Microsoft Edge Add-ons listing. Store availability and versions can differ. For other clients, use the Open VSX listing or the public release page.

2. Connect a model

Choose a provider and model in settings. Local choices such as Ollama and llama.cpp need a running local server. Hosted providers may need a provider-issued API key. For exact endpoint and data behavior, see Models and providers and the Privacy Policy.

3. Begin with a focused task

Open the correct workspace and ask for one clear outcome. Examples: “Run the existing test command and explain the first failure,” or “Update the mobile navigation and show the changed files.” A direct check can stay focused; multi-part product work can use a Scrum plan.

4. Review the result

Inspect changed-file diffs and artifacts. For implementation requests, check the validation output and any screenshot or browser evidence before approving or reporting completion.

The agent can make mistakes. Review generated code and commands, especially before running them on sensitive or production systems.

Install or update the terminal client

Run npm install --global @reallexi/rmai and confirm with rmai --version. The CLI has its own provider settings. See release notes and downloads.