withHuman

Meet withHuman

withHuman adds an approval step to actions performed by AI agents. It intercepts supported tool calls before execution and evaluates them against a configurable pipeline. Routine requests can be approved or denied automatically, while requests needing judgment are routed to an appropriate person or team with the action and its context.

Reviewers can receive notifications through their preferred connected channels, and escalation rules handle unanswered requests. Decisions and reviewer identities remain in an audit history. The open-source service supports coding-agent hooks, MCP workflows, and framework integrations, making it useful for teams that want agents to act while retaining oversight of selected operations.