RepoPatchBot
Ask me to build or fix something
Multi-Agent
Live URL of this repo's deployment. Used for verification + screenshots. Saved per repo — won't need to enter again.
analyse
standby
build
standby
review
standby
Pipeline retries automatically until verifier approves (up to 3×)
Merge PR
Set RepoPatchBot URL in Settings first, then come back here.
API Keys
Your deployed Railway service URL (RPBOT_URL).
Set OPENROUTER_API_KEY on Railway, or paste here to use directly.
Set GITHUB_TOKEN on Railway, or create a PAT with repo scope.
Deployment Platforms
Add tokens for platforms where you want the bot to deploy after every pipeline run. Static sites → GitHub Pages (no extra token needed, uses the token above). Frontend / Next.js → Vercel. Backend / bots → Railway.
VERCEL_TOKEN — for Vercel deployment (Next.js, React, static frontends). Get at vercel.com/account/tokens.
RAILWAY_TOKEN — for Railway deployment (Python, Node.js, bots). Get at railway.app/account/tokens. One-time: link your GitHub repo to Railway, then it auto-redeploys on every push.
From @BotFather. Pipeline auto-registers the Telegram webhook after deploy.
Semantic Code Memory (Qdrant)
When configured, the edit pipeline uses vector embeddings to retrieve only the relevant code chunks — reduces token usage and improves context quality.
From Qdrant Cloud dashboard. Leave empty for local/no-auth instances.
Built-in fastembed runs on your server (Railway). No extra cost, no API calls.
Balance & Credits
Enter the service URL above first.
Activate License Code
Enter a prepaid license code to add credits.
AI Assistant
Pipeline Models
Per-agent model for the 3-phase pipeline. Saved locally and sent with each run.
Breaks instruction into a plan
Implements the code changes
Reviews and validates the output