Skip to the collection

agent-native · Oversight & Approval

HumanLayer

Listed on The Agent-Native Index

Human in the Loop for AI Agents

What it does

HumanLayer is an API and SDK that lets AI agents request human approval before executing high-risk functions. The agent annotates specific functions with @require_approval(); when called, HumanLayer pauses execution, routes an approval request to a human via Slack, email, SMS, or other channels, and resumes the agent with the human’s decision injected into its context window.

The control flow is inverted from all traditional approval workflows: the agent initiates, the human responds, and the response feeds directly back into machine reasoning.


Agent-facing interfaces

Protocol surface

Interface Detail
Python SDK @humanlayer.require_approval() decorator; async-compatible
TypeScript SDK Equivalent TypeScript API
REST API Programmatic approval lifecycle management
Webhooks Receive human responses as webhook events
Slack Approval requests as interactive Slack messages
Email Approval requests via email with clickable approve/deny links
Discord Approval channel integration
Teams / SMS / RCS Advanced tier channels