@micdrop/basic

August 25, 2026 · View on GitHub

The smallest possible Micdrop app: a browser page with one button, a Node WebSocket server, and OpenAI for the three AI steps (agent, speech to text, text to speech).

Three files worth reading, around 150 lines in total:

src/server.ts runs the call and plugs in the AI providers
src/client.ts starts the call and displays its state
index.html holds the button and the conversation list

Run it

From the root of the repository, build the packages once:

pnpm install
pnpm build

Then add your OpenAI key:

cp .env.example .env

And start both the server and the page:

pnpm dev

➡️ http://localhost:8084

Click Start call, allow the microphone, and talk.

Where to go next

Getting started
Client documentation
Server documentation
Other AI providers, including models running on your own machine
React hooks for a real UI

For a full featured demo, with every provider wired in, see the advanced example.

License

MIT