withHuman
Human approval layer for AI agents built on an open protocol
Visit website →About
withHuman intercepts every tool call your AI agents make before it reaches the target system. Policy rules approve or deny routine calls automatically — refunds under twenty dollars, read-only queries, standard emails.
Anything that falls outside the rules goes to the right human reviewer, notified on Slack, Discord, SMS, WhatsApp or the mobile app.
Built on the open Agent Approval Protocol (AAP), it works with Claude Code, Codex, LangGraph, CrewAI, Pydantic AI, OpenAI Agents, Google ADK and Mastra.
Every decision lands in an append-only audit log with reviewer identity attached. The core is open source and self-hostable. Scale is twenty dollars per user per month with unlimited agents and approval requests.
AI-assisted draft, human-reviewed before publishing — see how we choose & review tools.
Why we picked it
As AI agents start making real payments, merging code and sending emails, someone needs to hold the door. withHuman built an open protocol for exactly this — not another approval dashboard, but a proper gate with escalation paths, audit trails and framework-agnostic hooks. The fact that it is open source and self-hostable makes it a serious piece of infrastructure.
Related tools
- 1 guilds.run Open-source AI agent teams you can actually inspect Free
- 2 OpenSwarm An AI-first OS where agent swarms do the work in parallel Free
- 3 Opengeni Open-source infrastructure for AI agents in your product Freemium
- 4 iFixAi Independent auditing for AI agents that catches hidden misalignment Freemium
- 5 Recurse Serverless harness that lets your agents compose specialists Freemium
- 6 Eclatira Conversational video AI that plugs into any application stack Freemium