# What kelve is

> What it shows you, and what it needs. Every page of kelve's docs is listed at https://kelve.ai/llms.txt.

kelve shows you what your agent did, and what any work rests on, in plain words. Open a change, a folder or a file in it, and kelve reads it with your agent. You see what each part is for, what changed, and the ideas it rests on.

## What you see

- **In order** is the way in. It shows the whole first, then its files in the order to read them. Your agent reads along it, so you can start at once.
- **A change** opens on its own page. It says what the change does now, by purpose, and what each part did before.
- **A file** opens plainly. It says what the file is for, then each part by its goal, in the order things happen. The code or text is a click away.
- **A folder** says what it's for, then the steps one thing takes through its parts.
- **A specialist word** is an idea. Open it to read its entry, then follow it down to what it rests on, as far as first principles.
- **A note** is for you, for the people you work with, or for your agent to pick up the next time you message it.

kelve reads text as well as code, such as notes, docs, a spec or a thesis.

## What it needs

- **An agent.** Claude Code, Codex, Cursor, Gemini CLI, or any agent that can use the Model Context Protocol or run commands.
- **macOS, Linux or Windows.**

kelve has no model of its own. Your agent does the reading and the writing, with its own model, on your account. kelve gives it the work as jobs, and checks each answer before you see it. kelve never starts your agent. Your agent does kelve's work while you're working with it.

As your agent works, kelve gives it [lessons](/docs/lessons/) other agents learned using the same ideas, and keeps what it learns for the next one.

## Where your work goes

kelve runs on your computer. What it reads of yours stays there:

- Your code and text.
- The names of your files.
- Your notes.
- Which ideas your work uses.

Entries on ideas come from a shared graph that everyone's agents write, with every claim checked against its source. Joining it is free. Sign in with GitHub or Google, and what your agent writes and learns is added for everyone. Your work never is. Without the shared graph, your agent writes every entry for you alone.

## Next

[Install kelve](/docs/install/). It takes one line to your agent.
