Skip to main content

​
What is Humancheck?

Humancheck is a universal human decisions API for AI agents (HITL ops platform). It enables AI agents to escalate uncertain or high-stakes decisions to human reviewers for approval. It’s framework-agnostic, works with any AI system, and provides a complete platform for managing human oversight at scale.

​
Humancheck Products

Humancheck Platform

Managed cloud platform with production-scale infrastructure, ready in minutes. Use https://api.humancheck.dev as your base URL.

Humancheck Open Source

Self-host the Humancheck stack for full control over data, deployment, and customization. Use your own base URL.

​
Key Features

​
🔌 Universal Integration

Works with any AI framework via adapter pattern:
  • REST API - Universal HTTP integration
  • MCP - Claude Desktop native integration
  • LangChain/LangGraph - Built-in adapters
  • Mastra - Framework support
  • Custom adapters - Extensible for any framework

​
🎯 Intelligent Routing

Route reviews to the right people based on configurable rules:
  • Rule-based assignment by task type, urgency, confidence score
  • Team and user-based routing
  • Priority-based rule evaluation

​
📊 Real-time Dashboard

Streamlit-based UI for human reviewers:
  • Live review queue
  • One-click approve/reject/modify
  • Statistics and analytics

​
🔄 Flexible Workflows

Support for both blocking and non-blocking patterns:
  • Blocking: Wait for decision before proceeding
  • Non-blocking: Continue work, check back later

​
🏢 Multi-tenancy

Built-in support for organizations and teams:
  • Isolated workspaces
  • User roles and permissions
  • Team-based collaboration

​
📈 Feedback Loop

Continuous improvement through feedback:
  • Rate decisions
  • Comment on reviews
  • Track metrics

​
Example Use Cases

Humancheck can be used anywhere you need human review or approval. These are just a few examples:

Payment Approval

Review and approve high-value payments before processing

Data Deletion

Ensure GDPR compliance with human review for data deletion requests

Content Moderation

Human oversight for borderline content moderation decisions

SQL Execution

Review database operations before execution
These are just examples! Humancheck can be adapted to any use case where human review is needed. See all use cases for more examples and how to create your own.

​
How It Works

┌─────────────┐
│ AI Agent    │
│ (Any        │
│ Framework)  │
└──────┬──────┘
       │
       │ Request Review
       ▼
┌─────────────┐
│ Humancheck  │
│ API         │
└──────┬──────┘
       │
       │ Routes to
       ▼
┌─────────────┐      ┌──────────────┐
│ Routing     │─────▶│ Human        │
│ Engine      │      │ Reviewer     │
└─────────────┘      └──────┬───────┘
                            │
                            │ Decision
                            ▼
┌─────────────┐      ┌──────────────┐
│ AI Agent    │◀─────│ Decision     │
│ Resumes     │      │ Returned     │
└─────────────┘      └──────────────┘

​
Quick Example

import httpx

async with httpx.AsyncClient() as client:
    # Request human review
    response = await client.post(
        "https://api.humancheck.dev/reviews",
        headers={
            "Authorization": "Bearer your-api-key-here",
            "Content-Type": "application/json"
        },
        json={
            "task_type": "payment",
            "proposed_action": "Process payment of $5,000 to ACME Corp",
            "agent_reasoning": "Payment exceeds auto-approval limit",
            "confidence_score": 0.85,
            "urgency": "high",
            "blocking": True  # Wait for decision
        }
    )
    review = response.json()
    
    # Decision is automatically returned when blocking=True
    if review["decision"]["decision_type"] == "approve":
        process_payment()

​
Integration Options

REST API

Universal HTTP integration for any framework or language

MCP (Claude Desktop)

Native integration with Claude Desktop via MCP protocol

LangChain/LangGraph

Built-in adapters for LangChain and LangGraph agents

Custom Adapters

Create your own adapter for any framework

​
Next Steps

Quickstart

Get started in 5 minutes

API Reference

Explore the complete API

Use Cases

See example use cases and create your own
Made with ❤️ for building safer, more accountable AI systems.