Kythera Docs
Kythera Docs

Welcome to Kythera

Kythera is an advanced AI Agent Builder platform designed to democratize the creation of complex, autonomous agent systems. Whether you need a simple chatbot, a multi-agent research team, or an automated HR assistant with database capabilities, Kythera provides the tools to build, test, and deploy in minutes.

psychology

Natural Language Builder

Describe your agent in plain English, and our Architect Agent will design the optimal workflow for you.

database

Built-in Persistence

Agents come with integrated Vector Search (RAG) and JSON Database capabilities out of the box.

Quick Start

Follow these simple steps to get your first agent running.

1

Create an Account

Sign up for a free account to access the workspace.

2

Generate an API Key

Go to the API Keys section in your dashboard and create a new key. You'll need this to test your agents in the playground.

3

Build an Agent

Head to the Workspace and chat with the Architect. Try saying: "Make a personal finance assistant that tracks my expenses."

Building Agents

Kythera uses a conversational approach to agent design. Instead of dragging nodes manually from scratch, you collaborate with our Architect Agent.

The Builder Workflow

  1. Requirement Analysis: You describe your goal. The system analyzes requirements.
  2. Architecture Design: The Architect proposes a graph structure (Agents, Tools, Routing).
  3. Generation: The system generates the configuration and visualizes it on the canvas.
  4. Refinement: You can ask for changes, e.g., "Add a web search tool" or "Make it more formal."

Testing in Playground

Once your agent is built, switch to the Playground tab to test it interactively.

Requirement

You must select an active API Key in the playground header to run the agent.

The playground supports real-time streaming, tool execution visualization, and session management.

Workflow Editor

You can fine-tune your agent by clicking on nodes in the visual canvas.

  • Agent Nodes: Edit the system prompt, model selection (Kimi, Gemini, GPT-4), and temperature.
  • Tool Nodes: Configure tool-specific settings like search limits or database schemas.

RAG / Knowledge Base

Give your agent long-term memory and access to private documents using Retrieval-Augmented Generation (RAG).

How to use:

  1. Click on a Vector Search or RAG tool node in the canvas.
  2. Open the configuration panel.
  3. Click Upload Document to add PDFs, text files, or Markdown.
  4. The system automatically chunks and embeds your documents.

Your agent can now answer questions based on these documents using the vector_search tool.

JSON Database

Kythera provides a built-in NoSQL-like database for every agent. This allows agents to remember structured data (user profiles, bookings, tasks) across sessions.

Data Viewer

You can view and manage the data created by your agents:

  1. Go to the Data page in the top navigation.
  2. Select your agent from the list.
  3. Browse collections and delete records if necessary.

Telegram Integration

Deploy your agent to the real world by connecting it to Telegram.

Setup Steps:

  1. Open Telegram and chat with @BotFather.
  2. Send command /newbot and follow instructions.
  3. Copy the HTTP API Token provided.
  4. In Kythera, go to your Agents page.
  5. Click the extension Integrations button on your agent card.
  6. Select Telegram, paste your token, and connect.

Your agent will now respond instantly to messages on Telegram!

API Access

Developers can integrate Kythera agents into their own applications using our Runner API.

Endpoint: Run Agent

POST https://kythera-agents-production.up.railway.app/run/stream

Headers

  • Content-Type: application/json
  • X-API-Key: YOUR_API_KEY

Body

{
  "agent_id": "agt_12345...",
  "message": "Hello world",
  "session_id": "optional-session-id"
}