For the complete documentation index, see llms.txt. This page is also available as Markdown.

Overview

Here is a cleaner Overview version for GitBook:

Overview

The Eitherway AI Companion is being built to behave less like a code generator and more like a working partner inside the platform.

That distinction matters.

A code generator takes a prompt, produces output, and hands the result back. The Eitherway AI Companion is being designed to do more than that. It is intended to plan before acting, explain what it is about to do, check whether its work actually succeeded, recover from errors where possible, remember project context over time, and operate within clear safety boundaries.

The goal is not just faster output. It is a system users can work with directly across the full lifecycle of what they build.

The companion is also being designed to work across the broader Eitherway experience, not only inside the main platform interface. As the product expands, users will be able to interact with it through messaging-first environments already central to the Eitherway workflow, including Telegram, with broader support being explored across additional interfaces. The aim is to let users work with the companion in the context that feels most natural to them, rather than forcing all interaction into a single surface.

The internal standard for the companion has been simple from the start: it should behave more like a teammate than a tool.

That means it does not just respond to prompts with isolated code. It is expected to reason through a task, break the work into steps, surface its plan in plain language, carry out the work, verify the result, and return with something the user can actually trust.

This is a different standard from traditional AI tooling.

It is also a more useful one.

Last updated