Agno: build a Discord alternative to OpenClaw

The presenter walks through a personal assistant built with Agno, a Python framework, as a customizable alternative to OpenClaw. The example runs on a VPS and accepts messages through Discord. It can inspect server processes, execute shell commands and edit files. An hourly test job gathers technology news and GitHub repository information, then posts a report to Discord.

The walkthrough focuses on the project's structure and configuration. It covers model selection, memory settings, tool access, workspace files and separate main and research agents. The presenter also explains how skills and scheduled jobs fit into the assistant. Agno's toolkit supplies integrations, with GitHub shown as an example that requires a package and token.

Setup uses uv, a Discord bot token and a model API key. The demonstrated model is Step 3.5 Flash through OpenRouter, which the speaker describes as free at the time. This is a self-hosted AI agent, but the example uses a remote model service; the video does not demonstrate offline inference.

The final section covers a user service that loads environment variables and starts the bot on the VPS. The presenter calls the project an early version and says Codex helped build it. Claims that this approach is lighter or potentially more secure than OpenClaw are the speaker's assessment, without comparative testing in the walkthrough.