Ordalin
Submit a tool

Website-checked tool profile

withHuman

Hold AI agents' risky tool calls for rules, an AI judge or a human reviewer on Slack, phone or email, with an append-only audit log.

Sources checked

Visit website ↗
Pricing
The open-source core is free to self-host. Hosted withHuman has a Free plan ($0, up to 5 members, limited AI judge), Scale at $20 per user per month with unlimited members, and Enterprise on a custom annual contract with on-prem or BYOC hosting. Agents and approval requests are unlimited on every plan. Official source 1 ↗ · Official source 2 ↗
Available as
Hosted web app at app.withhuman.ai or a self-hosted deployment of the open-source core. The local CLI connects coding agents on Mac or Linux. · Reviewers can approve requests from iPhone and Android apps with push notifications, as well as Slack, Discord, email, text message and WhatsApp. Official source 1 ↗ · Official source 2 ↗ Official source 1 ↗
Primary group
Automation & Agents

Also consider

See all 6 similar tools ↓

What it does

withHuman is an approval layer for AI agents. It intercepts an agent's tool calls, runs them through rules you write (with an optional AI judge or your own webhook), and sends the ones that need a person to the right reviewer on Slack, Discord, email, text, WhatsApp or its mobile apps, keeping an append-only audit log of every decision. The core is open source and self-hostable, with a hosted version run by Metoro Inc.

Official source 1 ↗ · Official source 2 ↗

Features

  • Tool-call interception. Each tool call is held by withHuman until it is decided. Hooks cover Claude Code, Codex and other coding agents, and SDKs cover LangGraph, CrewAI, Mastra and similar frameworks. Official source 1 ↗
  • Approval pipelines. Policy rules approve or deny routine calls automatically based on agent, tool, arguments and context; an AI judge handles calls rules can't describe, and webhooks can ask your own policy service. Official source 1 ↗
  • Escalation paths. Route requests to specific teams or people by agent, tool or arguments, with on-call order and escalation when nobody answers in time. Official source 1 ↗
  • Append-only audit log. Every request, decision and note is kept with the reviewer's identity. Official source 1 ↗
  • MCP gateway. Keeps an MCP server's credentials in withHuman instead of on agents' machines and sends every call through the approval pipeline. Official source 1 ↗
  • Open Agent Approval Protocol. Agents integrate against AAP, an open spec started by withHuman; the spec, SDKs, adapters and conformance tests are open source. Official source 1 ↗
  • Guided setup in plain language. The CLI detects installed agents, and you describe in your own words which calls need human review. Official source 1 ↗
  • REST API and MCP server. A REST API with an OpenAPI schema covers the review queue, agents, pipelines and settings, and a withHuman MCP server lets an AI agent configure pipelines, teams and webhooks. Official source 1 ↗ · Official source 2 ↗

Pricing & free limits

Open Source

$0self-hosted

The open-source core (queue, audit trail and web UI) on your own infrastructure. Hosted-only features such as the MCP gateway, AI judge, text/WhatsApp and mobile apps are marked not included in the comparison. Official source 1 ↗

Free

$0hosted

Up to 5 members, unlimited agents and approval requests, MCP gateway, text/WhatsApp and mobile-app notifications, audit log, a monthly allowance of AI judge evaluations, and email support. Official source 1 ↗ · Official source 2 ↗

Scale

$20per user / month

Unlimited members, full AI judge, audit log, organization and scoped API keys, single sign-on, verified domains, SCIM directory sync and email support. Seats are bought in advance. Official source 1 ↗ · Official source 2 ↗

Enterprise

Customannual contract

Adds on-prem or BYOC hosting, bring-your-own AI keys, SIEM export, a dedicated support engineer and 30-minute response for blocking issues. Official source 1 ↗

Billing details

Scale seats are purchased as a quantity; reducing seats below current use blocks new invitations rather than removing anyone. Billing runs through Stripe or AWS Marketplace. Official source 1 ↗ · Official source 2 ↗

A trial includes Scale features; without a subscription the organization moves to the Free plan when it ends. Official source 1 ↗

Free access & limits

Hosted Free allows up to 5 members and only a limited monthly AI judge allowance; organization and scoped API keys, single sign-on, verified domains and SCIM need Scale. Official source 1 ↗

Best for & limitations

Letting a support agent issue refunds while anything over a set amount waits for the payments team's approval. Official source 1 ↗

Putting coding agents such as Claude Code or Codex behind approvals for destructive commands, database writes or merges. Official source 1 ↗

Gating irreversible actions in finance, sales and marketing agents, such as paying bills, sending contracts or launching campaigns. Official source 1 ↗

  • Hosted features missing from open source. The self-hosted open-source edition lacks the MCP gateway, AI judge, text/WhatsApp and mobile notifications, audit log and SSO listed for hosted plans. Official source 1 ↗
  • Session recording on the hosted app. The hosted web app records sessions (pages, clicks, what is on screen, with typing hidden) through PostHog. Official source 1 ↗
  • Request data stays for the life of the organization. Approval requests, decisions and audit records, which can include customer data your agents send, are kept for as long as the organization exists. Official source 1 ↗
  • CLI setup is Mac and Linux only. The guided local setup requires a Mac or Linux machine. Official source 1 ↗
withHuman product screenshot
Website previewFixed desktop viewport · 1440 × 900

Similar tools

Published tools that share its categories.

PancakeAI go-to-market agent that finds buyers with live signals and sends approved, personal outreach from your account.
Marketing & SalesFree trialChecked Oct 15
Visit
MausBotOpen-source desktop app for running a team of AI agent bots on your own computer, using the Claude, Codex or Grok subscriptions you already have.
Automation & AgentsFree planChecked Oct 14
Visit
KaiAI executive assistant from Morgen that sorts Gmail and Outlook, takes bot-free meeting notes, turns commitments into tasks and plans your day.
ProductivityFree trialChecked Oct 13
Visit
guilds.runOpen-source, self-hosted framework for running teams of AI agents with any model, where every prompt, tool call and memory change is inspectable.
Automation & AgentsFreeChecked Oct 13
Visit