hey nekoDownload
free beta · windows & macOS

Your AI assistant and orchestrator.

Say “Hey Neko.” It answers out loud, remembers your clients and projects, and hands the building to Claude Code or Codex in the background, inside one folder it can never leave.

Free during beta · v0.2.0-beta36 · macOS on Apple Silicon or Windows · release notes

Thinks with your ChatGPT or Claude subscription, or Ollama on your machine. Builds with Claude Code, Codex or any CLI agent.

Neko

Speaking

● Microphone on (wake word processed locally)

  1. YouHey Neko, add dark mode to @awefolio, then commit and push it.

  2. NekoOn it. Claude Code is adding dark mode to awefolio. I'll tell you when it's pushed.

  3. Claude Code · awefolio · Editing globals.css · 1:12

  4. NekoClaude Code finished awefolio: dark mode is in, committed and pushed.

What is Hey Neko?

Hey Neko is a voice-first desktop assistant that coordinates the AI you already use: it talks with you, remembers your context, runs your local apps, and delegates coding work to agents in the background.

Local where it matters
The wake word and speech recognition run on your computer. Speech that isn’t “Hey Neko” is discarded on the spot.
Yours, not a provider’s
Name, voice, personality, memories and permissions live in Neko. Swap the model behind it any time.
Safe by default
It acts only inside your Apps folder, enforced by Neko’s own code rather than the model’s good intentions.
assistant + orchestrator

One voice in front. Your agents behind it.

Most AI tools are either something you talk to or something that does the work. Neko is both: an assistant you can talk to from across the room, and an orchestrator that keeps your coding agents and apps moving.

The assistant

  • Wake it by voice

    Say “Hey Neko” and talk normally. The wake word is detected on your computer.

  • Answers out loud

    A natural voice that runs locally, with 54 voices to choose from. Talk over it to interrupt.

  • Remembers you

    Memories you can see, edit and switch on and off, grouped by client, brand or project.

  • Knows your documents

    Add guides and specs to a memory group; Neko uses them while that group is on.

  • Live web and weather

    Web search, page reading and current weather when a question needs them.

  • Dictation and to-dos

    Hour-long recordings transcribed locally, plus a keyboard-first task outline.

The orchestrator

  • Delegates to coding agents

    Claude Code, Codex or any CLI agent builds and changes apps as background jobs.

  • Runs your apps

    Next.js, Vite, Expo and more: start, stop, restart and open them by name.

  • Ships by voice

    “Commit and push” is handed to the coding agent; force pushes and branch deletes are refused.

  • Survives restarts

    Unfinished coding tasks pick up where they left off the next time Neko starts.

  • Plugs into your tools

    Apps in your folder can expose MCP tools or Claude Code skills to Neko, on an allow-list.

  • Scriptable

    A token-protected local API on 127.0.0.1 for your own scripts and devices.

How a request flows

  1. 01

    Say it

    “Hey Neko, …” from across the room, or type in the chat. Neko hears the wake word on your machine.

  2. 02

    Neko routes it

    Quick things (time, timers, weather, your apps) are answered instantly. Everything else goes to the model you chose, and building work goes to your coding agent.

  3. 03

    Work happens in the background

    Agents keep working while you keep talking. Neko tells you, out loud, when something is ready.

who it's for

Built for four kinds of people who build with AI.

Neko is for individuals who use AI to make things on their own computer and are tired of being the glue between tools. If one of these describes you, it was designed for your day.

The multi-agent builder

You are: AI developers who already pay for Claude Code or Codex and keep several projects going at once on one machine, usually alone or in a team of two or three.

The problem: Your day is terminal tabs: one agent per project, dev servers left running, and the same context re-explained every time you switch.

  • Hand a task to Claude Code or Codex and keep talking; two jobs run in the background at once.
  • A live strip shows what each agent is doing (“Editing page.tsx”) and Neko tells you when it's done.
  • Start, stop, restart and open any app in your Apps folder without a terminal.
  • “Commit and push” goes to the coding agent, with force pushes always refused.

› “Hey Neko, add dark mode to @awefolio, then commit and push it.”

The fractional technical consultant

You are: Independent technical and AI consultants running three to six client engagements at the same time, each with its own codebase, vocabulary and rules.

The problem: Every client lives in a different head-space, and a general-purpose chatbot happily mixes them up.

  • A memory group per client: “switch to Acme mode” uses only Acme's memories plus your global ones.
  • Load a client's style guide or spec into its group as Knowledge (txt, md, PDF, Word).
  • Every client codebase sits inside one Apps folder that Neko's own code won't let it leave.
  • Dictate notes after a call: transcribed on your machine and copied to your clipboard.

› “Hey Neko, switch to Acme mode. What did we decide about the billing export?”

The marketer who ships

You are: Growth and content marketers on small teams who build their own landing pages, microsites and campaign tools with AI, and would rather not live in a terminal.

The problem: AI can write the page, but getting it built, running and published still means asking an engineer for help.

  • “Create an app called Spring Launch”: Neko makes the project and a coding agent builds it.
  • Keep each brand's voice and guidelines in its own memory group.
  • Neko starts the app and opens it in its own window, so you review it the way a visitor will.
  • Publish changes by asking: “commit and push spring launch”.

› “Hey Neko, create an app called Spring Launch: a one-page waitlist in our brand colours.”

The PM who prototypes

You are: Product managers and product-minded founders who test an idea with a working prototype before writing the ticket.

The problem: The prototype waits behind the sprint, and your best thinking is scattered across chats, docs and sticky notes.

  • Describe the prototype out loud; a coding agent builds it while you keep working.
  • A built-in task outline with Today, This Week, Waiting, Back Burner and Goals, which you can fill by voice.
  • “Start recording” captures a brainstorm; the transcript lands on your clipboard.
  • Mention an app with @ when typing, so Neko knows exactly which prototype you mean.

› “Hey Neko, add “interview three churned users” to this week.”

how it compares

Not another chat window. Not another terminal.

A chat app talks. A terminal agent builds. Neko sits between them and the rest of your machine, so you ask once and it works out where the request should go.

Hey Neko compared with a chat app and a coding agent in a terminal
You want to…Hey NekoAI chat appCoding agent in a terminal
Start a request hands-free“Hey Neko”, wake word detected on your computerOpen the app, then type or tap to talkType in the right terminal
Build or change an appHands it to Claude Code or Codex as a background jobWrites code for you to copyYes, one session per terminal
Run your local appsStart, stop and open them by nameNoYou run the commands yourself
Keep clients and brands apartMemory groups you switch on and offOne shared memoryPer-project files
Where it may actOnly your Apps folder, enforced by Neko's own codeDoesn't act on your filesDepends on the permission mode you pick
bring your own brain

Use the AI subscriptions you already pay for.

Neko doesn't resell model access. It signs in to the accounts you have, and with Ollama installed it falls back to a local model if your provider is down.

Conversation

  • ChatGPT subscription

    Default, through OpenAI's official Codex CLI. No API key needed.

  • Claude subscription

    Use your Claude plan as Neko's brain.

  • Ollama

    A local model, so nothing leaves your computer. Also the automatic fallback.

Coding

  • Claude Code

    Default coding agent, run headless inside the project folder.

  • Codex

    Sandboxed to the project's workspace.

  • Any CLI agent

    For example aider: put {task} where the task goes.

safety & privacy

Powerful on your computer. Fenced in by design.

Neko can create files and run commands, so its limits are enforced by its own code, not by asking the model nicely. Paths are resolved first, so “..” tricks and symlinks that point outside the Apps folder are caught.

What Neko runs, asks about, and refuses
RunsReading files in your Apps folder, creating project files, git status, npm test, starting a dev server
Asks firstDeleting or overwriting files, installing packages, unknown commands
RefusedAnything outside the Apps folder, .ssh/.aws/.env and keychains, credential dumping, curl | sh
  • Sleeping audio stays in memory. It’s never stored or uploaded, and transcripts of non-wake speech are never logged.
  • Your requests are transcribed locally. Only the text goes to the model you picked; with Ollama, nothing leaves your computer.
  • Conversations and memories live in a local database. History can be switched off.
  • Keys sit in your operating system’s credential store and are redacted from logs.
  • The local API listens only on 127.0.0.1, needs a token, and refuses browser origins.
questions

Frequently asked questions

What is Hey Neko?

Hey Neko is a voice-first AI assistant for Windows and macOS that lives in your system tray. You say “Hey Neko”, it answers out loud, remembers your context, runs your local apps, and hands building work to coding agents such as Claude Code and Codex in the background.

What does “AI orchestrator” mean here?

Neko doesn't try to be the only AI you use: it coordinates the ones you already pay for. It sends a conversation to your chosen model, delegates coding tasks to Claude Code, Codex or any CLI agent, tracks those jobs, and tells you when they're done.

Is Hey Neko free?

Yes, Neko is free while it's in beta. It uses an AI account you already have (a ChatGPT or Claude subscription) or a free local model through Ollama, so there is no separate API bill.

Who is Hey Neko for?

Hey Neko is for people who build with AI on their own computer: AI developers running several coding agents, independent technical consultants with multiple clients, marketers who ship their own pages, and product managers who prototype. It isn't a team product: there is no admin console or single sign-on.

Is Neko always listening, and is my audio uploaded?

The microphone is on, but only to hear the wake word, and that check happens on your computer. Speech that isn't “Hey Neko” is discarded immediately and never stored or uploaded. Your requests are transcribed locally too. Only the text goes to the model you chose, and with Ollama nothing leaves your computer.

What can Neko change on my computer?

Only what's inside your Apps folder. Neko's own code, not the AI model, blocks files, commands and paths outside it, refuses credential files such as .ssh, .aws and .env, and asks before deleting or overwriting anything.

Which coding agents does Neko work with?

Claude Code (the default), OpenAI Codex, or any command-line coding agent. Neko runs two coding tasks at a time, shows what each is doing, and resumes unfinished tasks after a restart.

Do I need to know how to code?

No. A setup wizard connects your AI account and microphone, and Neko starts and opens your apps without a terminal. Knowing your way around a project helps when you review what a coding agent built.

Which systems does it run on?

Windows, and macOS on Apple Silicon (M1 or newer). The beta isn't code-signed yet, so the first launch needs one extra click: “More info → Run anyway” on Windows, or “Open Anyway” under Privacy & Security on macOS.

Can I change its name, voice or language?

Yes. Pick an animal and a name and the wake phrase follows (“Hey Mochi”). Choose from 54 natural voices or your system's voices. Neko understands and answers in English and Korean.

Say the word. Neko runs the rest.

Free while in beta. Set up in a few minutes: the wizard connects your AI account, your Apps folder and your microphone.

Free during beta · v0.2.0-beta36 · macOS on Apple Silicon or Windows · release notes

  1. 01

    Download and open

    Drag Neko to Applications on a Mac, or run the installer on Windows.

  2. 02

    Allow it once

    The beta isn’t code-signed: “More info → Run anyway” on Windows, “Open Anyway” on macOS.

  3. 03

    Say “Hey Neko”

    Speech and voice models download in the background on first launch.