helderpgoncalves/feedbot

Feedbot is a self-hosted tool that converts feedback posted in Telegram groups into structured backlog items and lets AI agents or teams resolve them. Incoming messages are captured, optionally auto-triaged by an LLM (OpenAI/Anthropic) into fields like type, severity, summary and tags, and stored in a FastAPI + Postgres backend. Teams manage projects and invites via a React dashboard, issue MCP HTTP API keys, and use any MCP-compatible client (Claude, Cursor, Windsurf, etc.) to pick up and act on tickets. Resolutions and replies are posted back into the original chat; reporter replies reopen the ticket. Who uses it: engineering/product teams, community managers, and agent developers who want to route community feedback into workflows. Key features: Telegram integration, per-project LLM auto-triage, native MCP-over-HTTP server, secure defaults (encrypted keys, argon2id, HSTS), simple role model, frictionless onboarding, Docker/Coolify deployment, and LLM cost/audit controls.

AI-related keywords found in this repo's description, topics, or README summary — grouped by category. Each badge links to the corresponding ranking detail page.