Projects

Trading & finance

CPG Capital

An AI agent that sends me a daily portfolio briefing

· 2 min read

period
2026
role
Solo build
status
Running daily

Every morning an agent gathers data from several investment platforms into one view, explains what moved using the day’s news, and sends a short briefing to Telegram. Instead of a dashboard, I ask follow-up questions in the chat.

Long-term investing often means using several platforms, and no single app shows them together. CPG Capital pulls them into one picture.

A scheduled job runs each morning. It rebuilds an allocation report from each platform’s API or exported statements, writes a news digest with one-day, five-day and one-month moves and relevant headlines, and formats a short briefing with a “why it’s moving” section. A self-hosted AI agent delivers it to me on Telegram, and I can ask follow-up questions in the same chat.

The investment rules live in plain-text files that the agent reads, such as target allocations, so its advice stays consistent with the plan rather than reacting to headlines. Credentials stay in a local environment file that is never committed. I also prototyped CPG Wealth OS, a local-first web app version of the same idea for Australian investors, with platform connectors and rule-based rebalancing.

Highlights

  • Daily automated briefing delivered to Telegram
  • Combines data from several platforms through their APIs and exports
  • A news digest explains what moved and why
  • Rules-based allocation the agent follows

Built with

  • Python
  • PowerShell
  • OpenClaw agent
  • Telegram
  • MCP

What it taught me

The most useful automation is the one that turns up where you already are. A Telegram message gets read; a dashboard doesn’t.

The code for this project is private, so I’ve described how it works rather than linking to it. Nothing here is financial advice.