withHuman
A middleware product that provides a human approval layer for AI agents. It intercepts requests before tool invocations are executed, uses a rule engine to evaluate them, and routes approval requests to appropriate personnel, with cross-framework compatibility.
Timeline (last 90 days)
withHuman inserts a layer of human approval between AI Agents and the real world, letting rules handle routine operations and letting humans handle risky operations.
withHuman's workflow is divided into three steps: intercept tool calls, judge through a rules engine, and route high-risk calls to the corresponding owner in the team.
withHuman's rules engine supports three types of outcomes—allow list, deny list, or requiring human confirmation—with matching dimensions including tool name, parameter characteristics, call frequency, and the identity of the initiating Agent.
withHuman achieves cross-framework compatibility through the open protocol AAP (Agent Approval Protocol), and its official description explicitly mentions support for Claude Code, Codex, as well as Agents built with LangGraph, Pydantic AI, and other approaches.
withHuman adopts a combination of open source plus a hosted version; the open-source version can be self-deployed, code-audited, and customized as needed, while the hosted version lowers the barrier to entry.
withHuman received 80 votes and ranked 7th on Product Hunt, categorized under Open Source, Developer Tools, and Artificial Intelligence.
withHuman's maker provided a PRODUCTHUNT discount code for three months of free use of the Scale version.
All Facts (7)
withHuman inserts a layer of human approval between AI Agents and the real world, letting rules handle routine operations and letting humans handle risky operations.
50%UnverifiedwithHuman's workflow is divided into three steps: intercept tool calls, judge through a rules engine, and route high-risk calls to the corresponding owner in the team.
50%UnverifiedwithHuman's rules engine supports three types of outcomes—allow list, deny list, or requiring human confirmation—with matching dimensions including tool name, parameter characteristics, call frequency, and the identity of the initiating Agent.
50%UnverifiedwithHuman achieves cross-framework compatibility through the open protocol AAP (Agent Approval Protocol), and its official description explicitly mentions support for Claude Code, Codex, as well as Agents built with LangGraph, Pydantic AI, and other approaches.
50%UnverifiedwithHuman adopts a combination of open source plus a hosted version; the open-source version can be self-deployed, code-audited, and customized as needed, while the hosted version lowers the barrier to entry.
50%UnverifiedwithHuman received 80 votes and ranked 7th on Product Hunt, categorized under Open Source, Developer Tools, and Artificial Intelligence.
50%UnverifiedwithHuman's maker provided a PRODUCTHUNT discount code for three months of free use of the Scale version.
50%