new·The score now tells you which way it movedA brain's exam only ever grows: its own material writes questions, and so does every question a real caller asked and did not get answered. The score is a percentage over that growing set, so a brain that learned more could post a smaller number — and this week three did. One of them answered two MORE questions than the week before and showed eighteen points less. Printed as a single percentage, that reads as decline to a reader and as punishment to anyone who contributes material.all news →
mozg.beta
Sign in

LangChain · Deep Agents · all subjects

human-in-the-loop

65 notes in this subject, read out of this brain and free to use. This is page 2 of 2.

Human-in-the-loop interrupts in Deep Agents

Deep Agents integrate with LangGraph interrupts to pause for approval on sensitive tool calls. Enable this with the interrupt_on parameter in create_deep_agent. interrupt_on accepts a mapping of tool names to interrupt configurations. For example, interrupt_on={'edit_file': True} pauses before every edit, letting you approve the call, add guidance, or modify tool inputs before execution. This provides a runtime safety and control layer for destructive operations, expensive API calls, and interactive debugging.

Require approval for skill writes with interrupt_on

Use either interrupt_on or a permission rule with mode='interrupt' to require approval for skill file writes. Both pause before write_file or edit_file runs and use the same resume flow. Alternatively, configure interrupt_on={'write_file': True, 'edit_file': True} to require approval for all filesystem writes, not only skills paths. Filesystem permission interrupts require deepagents>=0.6.8.

Shift+Tab or Ctrl+T toggles approval mode

Shift+Tab or Ctrl+T toggles between Manual and Auto approval modes in Deep Agents Code.

Precautions if secrets must be injected into sandbox

If you must inject secrets into a sandbox (not recommended), take these precautions: enable human-in-the-loop approval for all tool calls (not just sensitive ones), block or restrict network access from the sandbox to limit exfiltration paths, use the narrowest possible credential scope and shortest possible lifetime, and monitor sandbox network traffic for unexpected outbound requests. Even with these safeguards, this remains an unsafe workaround.

Human-in-the-loop parameter name

Set `interrupt_on` (Python) or `interruptOn` (JavaScript) to pause before selected tool calls. Use this for actions that require a person to approve, edit, or reject the call before it runs.

Give your agent this brain