Getting Started

Quick start guide for Kite AIR Platform

Understanding the roles in Kite ecosystem

Service providers: who want to register their services for AI agent on Kite Agent App Store for monetization.
Agent builders: who want to develop agents with services on Kite Agent App Store with Kite SDK
Agent users: who interacts with agents to accomplish tasks, they use KitePass to authorize payment securely.

Note on the Current Version

Though we plan to split agent builder and agent user, in the current version of the platform, the workflow is unified. The agent builder is also the agent user. You will build an agent with the KiteSDK and securely provide your KitePass within your development environment to run it. We'll provide secure way to authenticate users' KitePass for agent builder's agent.

Current Access Status

Kite AIR is currently in limited access mode. We're gradually opening access to users and service providers through our waitlist and invitation system.

Account Authentication

We are using Particle Network for account and default wallet. You'll need to create account on Particle Network, and each login requires email verification code sent by [email protected].

Workflow

This guide will walk you through that unified workflow. We're actively working on features that will allow you to distribute agents to end-users who can provide their own KitePass through a user interface.

For service providers, Kite AI offers a platform to access demand for AI agents and explore monetization opportunities. Follow these steps to register your service.

Step0: Prepare your service as an API Endpoint

The service is registered as an API endpoint. Please ensure your API endpoint is prepared for use. If exists, authentication should be implemented via an API key included in the header.

Additionally, prepare an OpenAPI schema in JSON format for a subsequent step.

Step1: Activate your account by accepting invitation

As a group owner, you'll receive an invitation from Kite AI admin. Group members will receive invitations from the group owner (Note: this invitation has 24 hour limit). Please follow the email instructions to set up your account.

Step2: Register your service on Kite AI

On the Kite top screen, click "Go ->" on "Register your service", or use the menu at top right to click "Register Service"

Fill in the required information on the registration page:

API Endpoint URL: Include the complete path to the endpoint. If your service has multiple endpoints on the same host, register a new service for each endpoint individually.
Pricing Model:
- "By Volume" - Recommended if your service value is influenced by the input and output data volume (e.g., generative models).
- "By Call" - Suitable if the value is relatively constant

Pricing model is immutable

In the current version, we support the use case you will build an agent for yourself. We offer KitePass to pre-authorize payment with spending rules (as an agent user), and offer a Python SDK to build agents by using KitePass / services listed on Kite Agent App Store.

Step1: [As an agent user] Claim KitePass

Visit https://app.gokite.ai/, click "Go ->" on "Register your Agent for a KitePass", or use the menu at top right to click "Claim KitePass"

Then, name your pass and write description

Passport name is immutable

And configure spending rules

Frequency: Set for which time window you want to set budget
Budget: Amount of payment you allow for frequency you set
Target Provider: For which type of providers you want to set the budget

Now your passport is ready for use, copy API key for Step 4.

Step2: [As an agent builder] Find useful services

Visit https://app.gokite.ai/ to browse service, on service detail page, copy service ID

Step3: Choose your integration method

You have two options to integrate with Kite services:

Install gokite package

pip install gokite

Build your agent with SDK

The KiteSDK is designed to make building powerful, service-driven agents simple and secure. With just a few core methods, you can handle authentication, discover service capabilities, and execute tasks with pre-authorized payments.

Authentication

import os
from kite_sdk import KiteClient

# Initialize the client to authenticate your agent with KitePass
client = KiteClient(api_key=api_key)

This code securely authenticates your agent, giving it the authority to access and call services on your behalf.

Service information: Tell your agent how to use the service

service_info = client.get_service_info(service_id="unique-service-id")
service_info

Service Info:
Service ID: agent_LpzYVzWGzjGiZBI3HTm4ZCKj
Description: Demo weather API. It returns the current weather.
Service URL: https://demo-weather-service.vercel.app/api/weather

API Endpoints:

  GET /api/weather
    Summary: Get current weather
    Description: Returns the current weather for a static location (San Francisco)
    Responses:
      200: Weather response
        Response fields:
          - location (string)
          - weather (string)
          - temperature (string)
          - description (string)
      401: Unauthorized - invalid or missing API key
        Response fields:
          - error (string)

With this method you can easily instruct your LLM how to use this service.

Service calling
```
client.call_service(service_id, inputs)
```
This methods call service with pre-authorized payment by KitePass. Combining with service information you can easily declare agent tool.

The Kite MCP Server provides a universal integration layer that enables any MCP-compatible AI application (like Claude, ChatGPT, or other AI assistants) to access Kite's agent identity and settlement infrastructure without requiring code changes.

What is MCP?

Model Context Protocol (MCP) is an open standard that allows AI applications to connect to external tools and data sources. The Kite MCP Server acts as a bridge between MCP-compatible applications and Kite's services.

Integration

You can configure MCP on your MCP-compatible AI application with the following URL:

https://mcp.prod.gokite.ai/{api_key_value}/mcp

Note that api_key_value has the previx api_key_

Many clients allow MCP servers to be connected simply by providing a URL through a user interface.;

Here is a screenshot of the interface Claude, a desktop client, uses for connecting to remote MCP servers:

If you are asked to specify a transport type as part of the setup, choose Streamable HTTP.

Some clients may require a more manual type of configuration via a file in a format such as JSON. Here is an example of how you can connect to your MCP server to Claude using the claude_desktop_config.json file:

{
  "mcpServers": {
    "kite-passport-mcp": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.prod.gokite.ai/api_key_{api_key_value}/mcp"]
    }
}

The exact format used by different applications may vary, but the syntax shown above works with most clients requiring this type of manual configuration.

Benefits:

No Code Changes - Works with existing MCP-compatible applications
Universal Access - Any AI assistant can use Kite services
Secure Payments - Automatic settlement through KitePass
Service Discovery - Browse and use services from the Kite Agent App Store

Security Best Practices

API Key Protection: Treat your KitePass API key as a sensitive credential
Environment Variables: Store API keys in environment variables, not in code
Access Control: Regularly review and update spending limits
Service Validation: Verify service capabilities before integration

Next Steps

Ready to learn more about the Kite AIR Platform? Check out the comprehensive platform guide for detailed information about:

What is Kite AIR and its vision
Core components and architecture
Ecosystem roles and use cases
Future possibilities and roadmap

PreviousUse Cases NextKite AIR Platform

Last updated 1 day ago

Understanding the roles in Kite ecosystem

Note on Sign up and Log in

Current Access Status

Account Authentication

Workflow

Step0: Prepare your service as an API Endpoint

Step1: Activate your account by accepting invitation

Step2: Register your service on Kite AI

Step1: [As an agent user] Claim KitePass

Step2: [As an agent builder] Find useful services

Step3: Choose your integration method

Security Best Practices

Next Steps