# Your first look

> One change, read plainly, idea by idea. Every page of kelve's docs is listed at https://kelve.ai/llms.txt.

This page walks through one change your agent made, from what it does down to the ideas it rests on. It takes about ten minutes. To try it before you install kelve, use [the example on kelve.ai](/). It's the real reader, on a change an agent made to a payments service.

## Open what your agent did

Your agent shows you its work in kelve when it's done. Or ask it, in the terminal where it works:

```ask
Show me what you did in kelve.
```

kelve opens in a new browser tab on the change. The first time, kelve shows what the install set up first, and asks whether to join the shared graph. Then open the change from there.

Your agent's first job is finding a way in, which is the order to read the change. Then it reads along that way for you, file by file. While it reads, your agent's line in kelve's bar says how far it's got, such as "Claude Code reading 3 of 17". Each file shows as soon as it's read.

## Read in order

**In order**, at the left, is the way in. Its first step is **What this change does**, the change's own page. Then come its files, each with a line on why to read it now. A step's square is filled once your agent has read it, ready for you. It's hollow once you have. **Next** marks where to go on. Read as far as you need, and stop when you know enough. [Open anything](/docs/open/#in-order) says more.

## Read the change

The change's own page says, in order:

1. **What it changes**, and why, in a sentence or two, with what your agent said of it.
2. **Its kind**: it adds, fixes, reshapes, takes out or updates something, with what it did before and what it does now.
3. **How it works now**, by purpose, not by file. Each part is named by its goal, with what it did **Before** where the change touched it.
4. **What else it touches**, and its files in the order to read them.

A part's files are links. Open one to read it.

## Read a file plainly

A file opens **Plainly**. It says what the file is for first, then each part, in the order things happen. Each part has a goal of a few words and a sentence or two on what it does.

Parts the change touched are marked **New**, **Changed** or **Taken out**. The rest are **As before**. A changed part says what it did **Before**.

**Show the code** under a part opens the lines it explains, with what changed and the ideas on them. **As code**, on the file's name line, shows the file as code. It shows what changed, with the lines as before folded away.

## Open an idea

A specialist word is an idea, with a small square beside it. Read plainly, it's a word in the sentence that uses it. As code, the square is beside its line, and its name is in the margin, with a line on how this work uses it.

A red square is new in this change. A grey one was already here. A square is filled until you've read its entry, then hollow.

Click a square or a name. Its entry opens beside the work. The entry gives the key sentence first, then how it works, with its sources. Under it, **Rests on** lists the ideas it needs.

## Trace it down

Click an idea under **Rests on**. Its entry opens beside the one you were on, and the path in kelve's bar shows the way you came. Keep going until you reach a first principle. A first principle is a black square, on a line, with nothing beneath it.

If an entry didn't help, say so with **Didn't help me understand**, under it. The next agent that opens the idea makes it better.

## Leave a note

Select lines in a file, then **Note**, or press N. A note can be **Just you**, **With the work** for anyone who opens it, or **For Claude Code**, for your agent to pick up the next time you message it.

## Next

- [Open anything](/docs/open/): a folder, a branch, a commit, or text.
- [Read plainly](/docs/plainly/): how a page is put together, and why.
