Demo
See it in action.
The web app on the left, the toolbar popup on the right.
NimbleBuddy is a voice-first AI companion. Talk to Anna out loud — she answers in under a second, picks up where your last call left off, and looks things up on the live web with Nimbleway.
Everything happens by voice, in real time, with a personality that stays warm and wholesome.
Speak naturally and interrupt any time. Speech-to-text, LLM and a lifelike voice stream over a low-latency real-time network.
Tell Anna your name and she remembers what you talked about last time — stored per user and re-injected on your next call.
Ask about news, weather or anything recent. Anna calls Nimbleway's nimble_search tool over MCP and answers in a sentence or two.
Talk to Anna from the toolbar without opening a tab — a Manifest V3 popup on the same backend, memory and search.
The web app on the left, the toolbar popup on the right.
Live web search inside the product. Anna calls Nimbleway's MCP server whenever she needs fresh facts.
Used to design, build, debug and ship the app, the Chrome extension and this page.
Used during development of NimbleBuddy.
Used as a development environment while building NimbleBuddy.
You'll need real-time voice credentials (App ID + Certificate), and a Nimbleway API key for web search.
Clone, add your keys to server/.env.local, start both servers, open localhost:3000.
git clone https://github.com/vim719/NimbleBuddy cd NimbleBuddy bun run setup # add your voice App ID, App Certificate # and Nimbleway API key to server/.env.local bun run dev
With the backend running, build the popup (or unzip the download) and load it unpacked.
cd extension bun install && bun run build # chrome://extensions → Developer mode # → Load unpacked → extension/dist
Why no live demo link? Anna's backend holds the voice App Certificate and API keys, so it runs server-side rather than on GitHub Pages. Clone and run locally in about two minutes.