ComposioHQ/trustclaw — reverse-engineered prompt

Reverse engineered prompt

GitHub

Build me a self hosted personal AI assistant like TrustClaw.

I want a web app where I can sign up, chat with an AI assistant, connect my apps through OAuth, and ask it to actually do useful things for me, like work with Gmail, GitHub, Slack, Notion, Calendar, Drive, and similar tools. It should remember important details over time, keep long conversations manageable, and let me schedule recurring tasks so it can run while I’m away.

Please make it secure by design. Don’t ask users to paste raw API keys for their apps. Use OAuth based tool connections, keep action logs, and run tool actions in a sandboxed way instead of on my computer. I also want Telegram support so I can message the assistant from a bot, plus normal web chat.

Make it easy to deploy on Vercel, with Postgres vector memory, optional Redis rate limiting, username and password login, and a simple clean dashboard. Include clear environment variable examples and setup instructions. Look up current docs online if needed.

Want more depth? Deep Reverse