Review code from Claude Code, Codex or Cursor in a native Mac app. Open the diff, follow callers, and ask AI Reader about a change. Click its source references to read the surrounding code.
30-second recording: review a diff, ask AI Reader, and follow a source reference. English interface; no audio.
Works on local repositories
Answers link to source lines
Local changes, commits, branches
tea-notes-demoYENPO
ts recipe.ts×
src / recipe.ts /cupsFor
1234567
export function cupsFor(waterMl: number) { return Math.floor(waterMl / 200);}// Fictional teaching input: 550 ml.const cups = cupsFor(550);// cups is 2.
TypeScriptUTF-8 · Source preview
✳ AI Reader +
How many cups does this amount make?
ANSWER
Only complete cups count.
For this fictional recipe, each cup uses 200 ml. With 550 ml, dividing gives 2.75 and Math.floor returns 2.
recipe.ts:1–3
What happens with less than 200 ml?↗
ts label.ts×
src / label.ts /teaLabel
1234567
import { cupsFor } from './recipe';export function teaLabel(waterMl: number) { return cupsFor(waterMl) + ' cups';}// teaLabel(550) returns '2 cups'.
TypeScriptUTF-8 · Source preview
✳ AI Reader +
Where does the count in this label come from?
ANSWER
The label calls the recipe helper.
teaLabel passes waterMl to cupsFor, then puts the returned integer into a label. Read recipe.ts to check how that number is calculated.
label.ts:1–5
Which input would produce zero cups?↗
ts recipe.ts×
src / recipe.ts /cupsFor
123456
export function cupsFor(waterMl: number) { return Math.round(waterMl / 200);}// Proposed change in this teaching example.// cupsFor(550) now returns 3.
TypeScriptUTF-8 · Source preview
✳ AI Reader +
Would rounding preserve the result?
ANSWER
550 ml would become three cups.
Math.round(2.75) returns 3, but three cups need 600 ml in this fictional recipe. The input 550 distinguishes the proposed change from the original.
recipe.ts:1–3
Compare both functions with the same input.↗
Files, questions, and source referencesFictional code and answers, written for this illustration
01 /WHAT YOU CAN DO
Let the AI write, then see what it changed.
Diffs, the editor, search, and the AI Reader share one window. Local Changes shows everything that changed since your last commit, and when an answer mentions a line, you can open it right there.
Open the repo your AI works in
Open the DMG, drag Yenpo into Applications, and launch it. Choose the root of the repository your agent is working in. Local Changes shows what changed since your last commit, and you can read and search before setting up AI.
Pick one changed function and jump to its definition, its callers, and the tests that use it. Ask the AI Reader what the change does, then open the lines it cites and check the answer yourself.
Look at one commit, or compare two branches, such as the agent’s branch and main. Read the inline diff and ask the AI Reader to look for regressions. The comparison stays on the commits you picked, and it is read-only.
Say your agent just reworked the request client. Here’s one way to check it in Yenpo.
Open the changes
Open Local Changes, or compare the agent’s branch with main, and find the files that touch retries.
Ask what the change does
Ask the AI Reader which calls now retry, how many times, and what happens when the last attempt fails.
Open the references
Preview the lines the answer cites, then open the callers and tests the agent didn’t touch.
Run a check
Run the relevant test or command yourself to confirm the explanation. The AI Reader reads code but doesn’t run it.
Development preview: queued questions, steering and follow-up suggestions are described here for the development build. Availability in the download depends on your installed version.
No. Local Changes, commit and branch diffs, search, Markdown, and language-server navigation all work without one. Connect your own provider when you want to ask the AI Reader. Your provider controls model access and any usage charges.
Swift, Rust, C, C++, Objective-C, Python, JavaScript, TypeScript, and Go, each through its language server. Yenpo can install rust-analyzer for you; the other servers come from Xcode or your PATH. What works depends on the server and on how your project is set up.
Does my code stay on my Mac?
Browsing and search stay on your Mac. When you ask the AI Reader something, your question and the relevant code go to the model endpoint you set up, so check that provider’s terms before using restricted code. The AI Reader only reads unless you allow edits for a run.
Which Macs does Yenpo run on?
It runs on Apple silicon and Intel Macs with macOS 14 or later, and one download works on both. Choose Yenpo → Check for Updates… to see whether a newer version is available.
A cache investigation shows when to queue another question, steer an active task, manage pending questions, and choose a follow-up suggestion in Yenpo.