Human approval
A person approves sensitive or irreversible actions before the agent takes them.
Also called Human-in-the-loop, Human intervention
OpenAI, Anthropic, LangChain
The problem
Early in deployment, you cannot yet trust an agent with every action.
Human intervention is a critical safeguard enabling you to improve an agent’s real-world performance without compromising user experience. It’s especially important early in deployment, helping identify failures, uncover edge cases, and establish a robust evaluation cycle.
Therefore
High-risk actions: Actions that are sensitive, irreversible, or have high stakes should trigger human oversight until confidence in the agent’s reliability grows. Examples include canceling user orders, authorizing large refunds, or making payments.
One of the most common uses of interrupts is to pause before a critical action and ask for approval. For example, you might want to ask a human to approve an API call, a database change, or any other important decision.
In use
Claude Code
Claude Code runs on a permission-based model: by default, it's read-only, which means it asks for permission before making modifications or running any commands.
Where it breaks
Constant approval requests lead people to approve without looking.
Constantly clicking "approve" slows down development cycles and can lead to ‘approval fatigue’, where users might not pay close attention to what they're approving, and in turn making development less safe.
Sources
- A practical guide to building agentsOpenAI
Human intervention is a critical safeguard enabling you to improve an agent’s real-world performance without compromising user experience.
- Making Claude Code safer with sandboxingDavid Dworken and Oliver Weller-Davies, Anthropic, Oct 20, 2025
- InterruptsLangChain
- Human-in-the-loop - OpenAI Agents SDKOpenAI
Use the human-in-the-loop (HITL) flow to pause agent execution until a person approves or rejects sensitive tool calls.
Added Nov 14, 2026