Many useful agents do not wait for a chat message. They run on a schedule, react to a webhook or a file change, and keep working for minutes or hours. That needs a runtime around the agent: something that starts runs from triggers, records progress so a crash does not lose work, enforces limits, and reports health.
These guides start with what an autonomous agent runtime does and when you need one. They then cover the pieces: starting agents from webhooks and cron schedules, durable runs that resume after a crash without repeating side effects, and long-running MCP tasks handled as background jobs. The last guides take an agent to production: running, debugging and deploying a project with one command, and the operational checks to have in place first.
The code in these guides runs on Promptise Foundry, an open-source Python framework for AI agents and MCP servers.