Player not loading? Watch on YouTube
This tutorial builds an Agno AgentOS platform with a coding assistant, tests it locally in Docker, then deploys it to Railway. The presenter starts with a setup prompt from os.agno.com and uses the template's skills to configure the service and database. The web interface connects to the local instance at localhost:8000 before switching to the deployed endpoint.
The self-hosted platform provides a shared execution layer for agents accessed through REST, MCP and Slack. The presenter says its database stays in the operator's environment and explains control over access rules and data storage. The walkthrough does not establish offline operation or show a local LLM backend.
Authentication is a required deployment step: the deployed service initially fails until the presenter obtains a JWT verification key through os.agno.com and adds it to the production configuration. The coding agent then reports checks of authentication and the MCP and agent endpoints. Claude connects through a custom connector with a consent step and secret.
A news agent called Radar keeps a ledger of previously reported items to avoid repeating updates and uses Parallel MCP. Later examples cover conversation traces, user memory and simultaneous agent streams in one container. The presenter also explains simulation-based agent improvement and ends with Slack setup using an app manifest pointed at the AgentOS URL.
