Human-in-the-loop means an agent pauses before certain actions and waits for a person to approve, change or reject them. It is the simplest way to let an agent do real work, such as issuing a refund or changing a record, while keeping a person responsible for the decisions that matter.
The hard part is choosing what to gate. Ask for approval on everything and people start clicking yes without reading, which is worse than no gate at all. These guides cover the whole flow: adding approval gates to an agent's tool calls, deciding which actions need a person and which can run on their own so reviewers stay alert, and reaching approvers where they work with Slack messages and webhooks.
The code in these guides runs on Promptise Foundry, an open-source Python framework for AI agents and MCP servers.