Generate captions, optimize content, and enhance with custom images for maximum engagement.
46M visits/mo-10%#3 Image & Design
$20/moStarting price
About withHuman
withHuman provides an approval layer that intercepts consequential actions taken by AI agents before they execute. It holds these actions for review based on configurable rules, allowing automatic approval of routine requests or routing to the appropriate human reviewer when necessary. The system supports agents across different frameworks and coding environments through hooks and the open Agent Approval Protocol. Reviewers receive notifications on their preferred platforms, including Slack, Discord, email, text message, WhatsApp, and mobile apps, and can make decisions directly from those channels. Escalation paths can be customized to ensure the right person is notified based on factors like agent, tool, arguments, or time of day. The platform also includes features like approval pipelines, AI judges for complex decisions, and webhook blocks for integrating with existing systems.
Key features
- Intercepts all agent tool calls
- Configurable approval pipelines
- Automatic approval for routine requests
- AI judge for complex decisions
- Multi-platform reviewer notifications
- Custom escalation paths
- Webhook blocks for integration
- Open Agent Approval Protocol
Use cases
- Automating financial transactions with human oversight
- Reviewing high-risk database operations
- Managing agent-initiated refunds or payments
Pros
- Intercepts tool calls from multiple agent frameworks
- Supports automatic approvals based on configurable rules
- Notifies reviewers on their preferred communication platforms
- Open source Agent Approval Protocol for custom agent integration
- Customizable escalation paths for team-based review
Cons
- Free tier does not include hosting or reviewers
- Enterprise tier required for SIEM export
- No support for on-prem hosting in Free tier
- Limited AI judge functionality in Free tier
Frequently asked questions about withHuman
What does withHuman do?
withHuman acts as an approval layer for AI agents, intercepting consequential actions before they execute and holding them for review based on configurable rules. It supports automatic approval of routine requests or routes complex decisions to the appropriate human reviewer.
Who should use withHuman?
Teams deploying AI agents in production environments that require oversight for critical actions, such as refunds, database changes, or financial transactions, will benefit from withHuman's approval and escalation capabilities.
How does withHuman integrate with existing systems?
withHuman supports agents across different frameworks and coding environments through hooks and the open Agent Approval Protocol. It also offers webhook blocks for integrating with existing policy services or internal systems.
Can withHuman notify reviewers on multiple platforms?
Yes, reviewers receive notifications on their preferred platforms, including Slack, Discord, email, text message, WhatsApp, and mobile apps. Notification channels can be customized based on reviewer preferences.
How are escalation paths configured in withHuman?
Escalation paths can be customized to ensure the right person is notified based on factors like agent, tool, arguments, or time of day. Teams can set their own escalation order and on-call flows.
What is the Agent Approval Protocol (AAP)?
The Agent Approval Protocol is an open standard that enables withHuman to intercept tool calls from AI agents regardless of the framework. It ensures compatibility with agents built on LangGraph, CrewAI, Mastra, and other platforms.
