W

withHuman

The approval layer for AI agents.

Open sourceBuilt on open standardsFramework-agnosticAgent Approval ProtocolWorks with Claude Code & Codex
Category
AI agent governance / human-in-the-loop approval layer
Stage
Early โ€” launching today according to landing page
Target audience
Teams running AI agents for consequential operations, including support, finance, SRE, sales, marketing, and coding agents.
Opportunity score
68/100

A promising developer-focused infrastructure concept in a growing AI agent market, with an unproven business model and meaningful risk of platform absorption.

Founder verdict
MAYBE

A well-positioned early product in a real and growing layer, but too unproven and platform-exposed for a broad standalone bet.

What is withHuman?

withHuman is an open-source, framework-agnostic approval layer that intercepts consequential AI agent tool calls and routes them through automated rules and human approval workflows. The landing page focuses on high-stakes actions such as refunds, payments, database writes, and production rollouts, emphasizing approval pipelines, escalation paths, notification channels, and append-only audit logs. The product is positioned as a lightweight layer between agents and tools rather than an agent builder. While the problem is clear and growing, revenue, growth, pricing, and customer proof are not publicly disclosed.

Get the full withHuman playbook

The complete reverse-engineering โ€” what to copy, what to avoid, and exactly what to build instead.

Unlock full startup intelligence โ€” $39.99