Decibot

Decibot records Discord voice channels, transcribes each speaker separately, and delivers clean Markdown transcripts to your server, built for teams that meet in Discord and for the AI workflows that come after the meeting.

The problem

Discord is where teams, communities, and founders actually talk, and everything said in a voice channel evaporates the moment it ends. Meeting notes get reconstructed from memory, decisions get lost, and none of it is usable by the AI tools teams increasingly run their work through.

What we built

A capture pipeline that subscribes to each participant’s voice stream individually, giving clean speaker separation without diarization guesswork. Utterances are transcribed with Whisper and stitched back into chronological order, producing a transcript that reads like the conversation actually happened.

Around the pipeline: eleven slash commands with a persistent session control panel, a free tier with Stripe-backed subscriptions, and an account portal with Discord sign-in, usage charts, and self-serve data export. The part we like most: Decibot is also an MCP server. Subscribers mint API keys and connect Claude, Cursor, or any MCP client, letting an agent join a call, record it, and pull the transcript into its own context.

Where it stands

Publicly launched July 17, 2026. Verified Discord app, live in the Discord App Directory, real billing from day one, and a full ops layer: error monitoring, uptime checks on a public status page, and daily backups.