Open-source human approval layer that puts high-risk AI agent tool calls behind human review.
Homepage hero - the approval layer for AI agents
1 / 2Overview
What is withHuman?
A open-source human approval layer for AI Agents
withHuman is an open-source human approval layer for AI agents. It intercepts every agent tool call through the Agent Approval Protocol before deciding whether it executes, with SDKs for LangGraph, CrewAI and Mastra plus hooks for Claude Code and Codex. Policy rules auto-approve or deny routine requests to eliminate approval fatigue, while escalation paths route the rest by agent, tool, params and team with on-call flows. Reviewers are reached where they already are: Slack, Discord, mobile push, email, SMS and WhatsApp. An append-only audit log records reviewer identity and the full request for every decision. The core is open source and self-hostable; the hosted Free plan covers up to 5 members with unlimited agents and approvals, and Scale costs $20 per user per month.
Platforms and languages
withHuman Availability
Platforms
Languages
Capabilities
withHuman Key Features
Tool-call interception
Intercepts every agent tool call before deciding execution via the Agent Approval Protocol; SDKs for LangGraph, CrewAI and Mastra, hooks for Claude Code and Codex.
No approval fatigue
Policy rules auto-approve or deny routine requests based on agent, tool, params and context; AI judge and webhooks to your own policy service.
Escalation paths
Route approvals by agent, tool, params and team; teams set escalation order and on-call flow; auto-escalates when no one responds.
Reviewers where they are
Reach approvers on Slack, Discord, iPhone/Android push, email, SMS and WhatsApp.
Audit log
Append-only history with reviewer identity per decision and the full request retained per decision.
Best for
Who uses withHuman?
Teams running AI agents in production
Require human approval for high-consequence agent actions in support, finance, SRE and coding
Plans and access
withHuman Pricing
Freemium
Free self-hosted & hosted Free $0; Scale $20/user/mo
Common questions
withHuman FAQs
How much does it cost?
The open-source core is free to self-host; hosted plans include a free tier for up to 5 members, Scale at $20 per user per month, and custom Enterprise contracts.
Is it secure?
It is SOC 2 Type II compliant and keeps an append-only audit log of every decision.
Can I self-host it?
Yes. The core is open source and designed for self-hosting.
Which agents does it work with?
It is framework-agnostic: SDKs for LangGraph, CrewAI and Mastra, hooks for Claude Code and Codex, an MCP gateway and an open protocol.
How do reviewers get notified?
Via Slack, Discord, iPhone/Android push notifications, email, SMS and WhatsApp.
Reviews
See what the community thinks and share your experience.
—
Based on 0 ratings
Rating distribution
Community reviews
No reviews
No written reviews yet.
