KirokuForms

Your agent needs a person. They do not need an account.

Send a review to anyone: a client, an accountant, the colleague who signs off deliveries. They get a link, answer the question, and your agent carries on. No signup, no password, no seat to buy for the person doing the reviewing.

It runs on an MCP server, native LangGraph tools and a Python SDK, so the request is a few lines from wherever your agent already lives.

Reviewers need no account MCP Compliant LangGraph Native Python SDK Dynamic Review Forms Full audit trail

Integrate Human Insight Your Way

KirokuForms offers flexible integration patterns to fit your existing AI architecture and development practices.

Use KirokuForms as an MCP Server

Implement standardized Human-in-the-Loop by connecting your AI systems to KirokuForms via the Model Context Protocol. Submit tasks, define review forms dynamically, and receive structured feedback. Every task lands as a case with an owner, a status, and an activity trail, so a person can pick it up and see it through to an outcome. See lightweight case management.

  • Fully MCP v0.2 compliant (or target version).
  • Handles task queuing, assignment (future), and reviewer UI.
  • One authenticated endpoint for every review your agents raise.
Full MCP Guide
Connect an agent (Python)
from kirokuforms_mcp import MCPClient

client = MCPClient(api_key="YOUR_API_KEY")

# Create a simple review task
task = client.create_task(
    title="Review Generated Content",
    fields=[
        {"name": "content", "type": "textarea", "label": "Generated Text"},
        {"name": "approved", "type": "radio", "label": "Approve?", 
         "options": [{"label": "Yes", "value": "yes"}, {"label": "No", "value": "no"}]}
    ]
)

print(f"Review at: {task['formUrl']}")

Native LangGraph Tool/Node

For LangGraph users, KirokuForms offers (or will offer) a dedicated node to easily insert human review checkpoints into your agentic workflows. Simplify `interrupt()` handling and manage human feedback loops.

  • Drop-in component for LangGraph agents.
  • Manages state and callbacks for human review.
  • Uses KirokuForms' dynamic form capabilities for review UIs.
LangGraph Integration Details
Pause a LangGraph run for a human (Python)
from langgraph.graph import StateGraph
from kirokuforms_langgraph import KirokuFormsHITLNode

# Add human review to your LangGraph workflow
workflow = StateGraph(AgentState)
workflow.add_node("agent_step", your_agent_logic)
workflow.add_node("human_review", KirokuFormsHITLNode(
    api_key="YOUR_API_KEY",
    form_template="review_template"
))

# Route to human review when needed
workflow.add_conditional_edges(
    "agent_step",
    lambda state: "human_review" if state.get('needs_review') else END
)

Convenient Python SDK

Our Python SDK (`kirokuforms` or `kirokuforms-hitl`) provides a high-level interface to interact with all KirokuForms HITL services, including task submission, dynamic form definition, and results retrieval.

  • Easy installation via pip.
  • Simplifies authentication and API calls.
  • Built for rapid integration into your Python-based AI projects.
Python SDK Documentation
Create a task with the Python SDK
from kirokuforms import KirokuClient

client = KirokuClient(api_key="YOUR_API_KEY")

# Create a dynamic review form
task = client.hitl.create_task(
    title="Content Moderation",
    fields=[
        {"name": "content_id", "type": "text", "label": "Content ID"},
        {"name": "action", "type": "select", "label": "Action", 
         "options": ["Approve", "Reject", "Escalate"]},
        {"name": "notes", "type": "textarea", "label": "Notes"}
    ]
)

print(f"Task ID: {task.id}")

Overcome Your Human-in-the-Loop Hurdles

KirokuForms addresses common pain points in implementing and scaling human review for AI.

Custom HITL UI Development

The Problem

Building and maintaining bespoke UIs for human review tasks is a significant engineering overhead, diverting focus from core AI model development.

Our Solution

Offload UI/UX to KirokuForms' pre-built, dynamic, and user-friendly review interfaces. Define forms via API or templates; we render and manage the review experience.

Key Benefits

  • Reduce frontend development time
  • Ensure consistent reviewer UX
  • Focus entirely on AI logic

Costly & Complex Platforms

The Problem

Traditional enterprise HITL platforms can be prohibitively expensive, slow to implement, and lack the flexibility needed for agile AI teams or research.

Our Solution

KirokuForms offers a developer-friendly, self-serve model with transparent and competitive pricing. Get started quickly and scale HITL affordably as your needs grow.

Key Benefits

  • Lower total cost of ownership
  • Faster time-to-value for HITL
  • Flexible plans for all team sizes

Lack of Standardization

The Problem

Ad-hoc HITL solutions for different AI projects lead to brittle, hard-to-maintain workflows and prevent reusability of review processes.

Our Solution

One protocol and one API for human review, so the same review process serves every project instead of being rebuilt per agent.

Key Benefits

  • Improved interoperability
  • Consistent review processes
  • Easier scaling of HITL operations

Powering Diverse AI Applications

From validating outputs to collecting nuanced feedback, KirokuForms supports a wide range of HITL needs.

AI Model Output Validation

Content Moderation Review

Complex Data Annotation

Exception Handling in Agents

Human Feedback for RLHF

Research Data Verification

Ready to Integrate Intelligent Human Oversight into Your AI?

Streamline your Human-in-the-Loop processes with KirokuForms. Explore our documentation or talk to our team about your specific HITL needs.

Or read the developer docs, explore enterprise HITL, or see pricing for AI features.