Powered by AssemblyAI
Stop typing prompts. Just ask.
Hold one key in VS Code and ask out loud. EchoCode answers in plain language, opens the file the answer is about — even one you don't have open — and lights up each line as it says it.
Hold Ctrl + Alt + Space and speak — on macOS, Ctrl + Shift + Space. Free, and no API key to set up.
9static List<Integer> findDuplicates(List<Integer> ids) {10 List<Integer> seen = new ArrayList<>();11 List<Integer> duplicates = new ArrayList<>();12 for (Integer id : ids) {13 if (seen.contains(id)) {14 if (!duplicates.contains(id)) {15 duplicates.add(id);
EchoCode
The problem
Many students now "vibe code": they tell an AI "please fix the code" or "add this feature" and paste in whatever comes back, without understanding what was wrong, why the fix works, or the logic behind the new code. Asking for a real explanation takes more effort, because it means typing a long prompt, copy-pasting code into a chat window, and matching the answer back to the right lines. So they skip the understanding and ship code they cannot explain. EchoCode makes asking why as easy as turning to the person next to you, so understanding the code takes less effort than copying it.
How it works
Four steps, and none of them leave your editor.
Wake it
Press the hotkey once. EchoCode says hello and waits. Until that press it is connected to nothing and your microphone is off.
Ask out loud
Hold the same key and ask the way you would ask the person next to you. Let go to send. Select some code first if your question is about one part of it.
Listen and look
The answer comes back in plain language. The file it is about opens, and each line lights up as it is mentioned. Ask to be taken somewhere instead — "open the LinkedList file" — and it comes up with your cursor in it.
Apply, or ask again
A suggested fix appears as a card you can insert with one click. Changed your mind mid-answer? Press the key again and ask something else.
What makes it different
Plenty of tools can answer a question about code. The difference here is everything you don't have to do first — and that the answer comes back pointing at your own screen.
You talk instead of typing
Hold one key and ask. No prompt to write, no code to paste, no window to switch to.
It already has your code
The file you are in, your cursor, whatever you selected and any errors in that file go with every question automatically.
It looks past the file you are in
Ask "where do I change the heading?" while looking at index.html and it finds the rule in style.css, opens that file and highlights the line. In our own testing on a small web app, 9 of 9 questions landed on the right file and line.
It takes you to the file
Say "open the globals.css file" and it opens with your cursor in it. A slightly misheard name like "global.css" still finds it, and if two files could match, EchoCode asks which one you meant.
It points while it talks
Lines light up in your editor at the moment they are spoken, so you never have to work out which part of your screen the answer means.
It explains, in everyday words
Two to four short sentences, and a technical term is explained the first time it is used — "a null pointer exception, which just means the code reached for something that isn’t there".
The fix is one click away
When the answer proposes a change, the exact code appears as a card. Replace the lines it is about, or insert it at your cursor.
Built on AssemblyAI
One streaming session runs the whole conversation, and your audio goes straight to AssemblyAI — our server never sits in the audio path.
Voice Agent API
One streaming session hears the question, reasons over your code and speaks the answer. Coding words — HashSet, generics, null pointer — are sent as key terms so they are recognised, and turn detection works out when you have finished talking.
Word-level timing
Each word of the answer comes back with its position in the spoken audio. That is what lets the right line light up at the moment it is said, instead of everything highlighting at once when the answer ends.
LLM Gateway
A second, separate call writes the code behind a suggested fix as JSON. It starts while the spoken answer is still playing, so building a code card never slows the conversation down.
Single-use session tokens
The AssemblyAI key lives only on our Vercel backend, which mints a short-lived token for each conversation. Audio then streams straight from your editor to AssemblyAI, with no server in the middle to slow it down.
What happens to your code
- Your microphone is off until you hold the key
- No AssemblyAI key is shipped inside the extension
- node_modules, .git, build output and lock files are never sent
- Files git ignores, and files named like a secret (credentials.json, .env, key files), stay on your machine unless it is the file you have open
- Your question and your code go to AssemblyAI to be answered. The code-card request passes through our Vercel server on the way there
- All we keep is an anonymous install id, the minutes you have used, and request counts per IP address for up to a day, to count your free minutes and limit abuse
Pricing
Free forever for students, with 10 minutes of voice every month. Bootcamps and universities pay for a larger allowance for the students they teach.
Free
Try EchoCode
$0
Available now. No card, no API key.
Download the extension- 10 minutes of voice a month, about 13 questions*
- Ask out loud and hear the answer
- Answers that look across your project, not just the open file
- Ask for any file by name and it opens
- Lines light up as they are mentioned
- One-click code cards
Pro
For everyday learning and coding
$20
Per month, or $192 a year — two months free.
Everything in Free, plus:
- 60 questions a month, about 45 minutes of voice*
- Planned: top up when you need more
- Planned: a longer warm session, so follow-ups stay instant
- Planned: a stronger model behind code cards
Cohort
For bootcamps and training programmes
$25
Per learner, once, for a 12-week cohort.
Everything in Free, for every learner:
- A larger voice allowance, pooled across the whole cohort
- No sign-up for learners — one invoice, everyone in
- Planned: progress view of where a cohort gets stuck
- Planned: overage by the minute, never a hard stop
Campus
For university departments
$36
Per seat, per academic year.
Everything in Free, for every student:
- A larger voice allowance, pooled across the department
- Priced per enrolled seat, billed once a year
- Planned: instructor view of the concepts a class keeps missing
- Planned: shared allowance across courses
*Minutes count the audio of your questions and EchoCode's answers; question counts are estimates. Only Free is available today. Pro, Cohort and Campus are not on sale yet, and the items marked planned are not built.
Ask your first question out loud
Install it, press the key, and ask what that method actually does. It takes about a minute.