I built a radio station with one listener, and it runs in a terminal
The film above is 86 seconds, and the middle of it is a real on-air moment. There is a longer unedited recording at the top of the repo README . Watch either first, with the sound on. The rest of this post is why it exists and how it is put together. What it is murmur runs in a terminal window and behaves like a radio station whose only listener is you. It picks a topic on its own and talks about…
The story is about a unique AI radio station called "murmur" that operates entirely in a terminal window. The radio station, which was built by the author, behaves like a radio station with only one listener - the user. It autonomously selects a topic to discuss, speaks in a human-like voice, and plays a song before returning to the program. At certain hours, it also announces greetings like good morning and good night.
The user does not need to respond to any of the radio station's dialogue. If the user types something, the AI host replies in character. Over time, the AI builds a separate profile of the user, incorporating their preferences and characteristics discovered during their interactions.
The AI behind murmur is powered by Claude, an AI agent, and runs on Node 24, without any build steps. It uses a small set of tools to perform tasks such as searching for songs, selecting the best candidates, committing a pick, updating memory, and running the setup guide. The AI's voice is sourced from a hosted HTTP endpoint, and it does not integrate with the user's existing Claude Code setup.
The radio station's logic, including its personality and memory, is stored on the user's machine as separate components. The user can modify the AI's character by editing a markdown file under the ~/.murmur directory. The AI's music selection is handled by yt-dlp, and its voice generation relies on a hosted service.
The story highlights the author's motivation behind creating murmur: to provide a more personal and engaging AI experience, separate from the task-oriented AI tools available. The author experimented with an evolving AI character, but found that it drifted towards a generic assistant voice, rendering the unique interaction unrecognizable. Therefore, the AI's character is now frozen and can be modified manually by the user.
One limitation of murmur is that it requires a Claude Code subscription for the AI brain, and it currently relies on hosted services for its voice and music selection. The author has plans to improve the onboarding process and make the software compatible with Linux and Windows, but as of now, it is primarily developed on macOS.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
