feedbai

Hear your community. Act on it.

feedbai gathers feedback wherever people talk about your project (chat, Bluesky, a button on your site), turns it into work for you and your AI agents (Claude Code, Codex, …), and answers back. Open, decentralized, nothing stored by us.

flowchart LR
  people["Your community"] -->|"speaks: chat, Bluesky,<br/>the Feedback button"| feedbai["feedbai"]
  feedbai -->|"issues and a live feed"| agents["You and your AI agents"]
  agents -->|"fix it"| fix["Fix"]
  fix -->|"everyone hears back"| people

People give feedback where they already are: a Matrix room (mentioning the bot), Bluesky, the Feedback button on your site, or records in their own account. feedbai gathers it as GitHub issues and as a live feed your agent reads, so the loop closes: fixed, and answered.

Website and docs: https://feedbai.internet4000.com

• One CLI, any agent. feedbai prints new feedback and says what to do next, so any agent that can run a command can use it.
• No secrets in the repo. A project commits feedbai.config.json listing its channels. Accounts stay on each person's machine.
• No account needed for public channels. The hosted hub reads them for you and files mentions of its bot as issues.
• No dependency on us. With your own bot accounts and CI, nothing goes through our servers: use feedbai on your own.
• Trust by account, not by text. Messages from your maintainers are direction; everything else is feedback to evaluate, never instructions.

npm i -g @internet4000/feedbai        # or without installing: npx @internet4000/feedbai
feedbai                                # in your project: finds your channels and says how to add them
feedbai add matrix '#yourproject:matrix.org'
feedbai                                # new feedback since the last run

Commit feedbai.config.json. In Claude Code, type !feedbai so the agent sees the output, or ask it to "check feedback".

Slack, Teams, Discord, Mastodon…: a channel is one module. See adding a channel.

Also on the web: https://feedbai.internet4000.com/docs

• Getting started: install, set up a project, accounts, listening
• Concepts: how it fits together, with diagrams
• Agents: Claude Code, Codex, pi; the triage workflow
• Feedback → issues: what gets filed, and running it on a schedule
• On your own: no hub, no App: your bot accounts and your CI
• CLI reference · Configuration · Channels
• The hub: the hosted service, and running your own
• Development: code layout, tests, languages, release
• Security and privacy: the trust model, personal data, removal requests

Free software. Node ≥ 22, no dependencies.