​
Kubiya Platform Overview

Kubiya is the first LLM-native automation platform that runs entirely on your infrastructure. Unlike traditional automation platforms, Kubiya is designed from the ground up for AI agents with Serverless Container Tools, LLM-Friendly DAG Workflows, and Zero-Trust Security - all executing in your own environment.

​
🎯 What Makes Kubiya Unique

🧠 LLM-Native Design

Built for AI agents - Every tool and workflow is designed to be easily understood and executed by LLMs

🏠 Your Infrastructure

Runs on your infra - Complete control and security with self-hosted runners in your own environment

📊 LLM-Friendly DAG Workflows

Visual workflows that LLMs can understand, modify, and execute with natural language descriptions

🛠️ True Serverless Tools

Container-native tools that scale to zero, start instantly, and run any language or framework

​
🏗️ LLM-Native Architecture

​
🔑 Key Architecture Principles

  1. 🏠 Your Infrastructure First: All execution happens in your environment
  2. 🧠 LLM-Native: Every component designed for AI agent interaction
  3. 📊 Minimal Data Transfer: Only metadata and telemetry leave your infrastructure
  4. 🛡️ Zero-Trust Security: Policy validation before any execution
  5. ⚡ Serverless Execution: Tools scale to zero when not in use

​
🎯 Core Capabilities

​
1. 🛠️ True Serverless Container Tools

What makes them special: Unlike traditional serverless functions with cold starts and language limitations, Kubiya’s serverless tools are container-native and LLM-optimized. 🎯 LLM-Optimized Features:
  • 📝 Natural Language Descriptions: Each tool has LLM-friendly documentation
  • 🔍 Semantic Discovery: LLMs can find tools by describing what they need
  • ⚡ Instant Scaling: From 0 to 1000+ containers in seconds
  • 🏠 Your Infrastructure: Run on your Kubernetes, Docker, or VMs
  • 📊 Live Streaming: Real-time output for LLM feedback
Example Tool Definition: 
{
  "name": "analyze-logs",
  "description": "Analyze application logs for errors and patterns",
  "llm_prompt": "Use this tool when you need to investigate application issues, find error patterns, or analyze log data. It supports JSON, text, and structured logs.",
  "image": "python:3.11-slim",
  "packages": ["pandas", "numpy", "matplotlib"],
  "integrations": ["aws/s3", "elasticsearch"],
  "scaling": {
    "min_instances": 0,
    "max_instances": 100,
    "scale_to_zero_timeout": "5m"
  }
}

​
2. 📊 LLM-Friendly DAG Workflows

What makes them special: Traditional workflows are code-heavy and hard for LLMs to understand. Kubiya workflows are declarative, visual, and LLM-optimized. 🧠 LLM-Optimized Features:
  • 📝 Natural Language Steps: Each step has human-readable descriptions
  • 🎯 Intent-Based: Focus on “what” not “how”
  • 🔍 Self-Documenting: Workflows explain themselves to LLMs
  • 🔄 Dynamic: LLMs can modify workflows on-the-fly
  • 📊 Visual: Mermaid diagrams auto-generated for LLM understanding
Example: LLM Creating a Workflow 
# LLM Request: "Create a data pipeline that processes user events"
# Kubiya generates:

workflow = {
    "name": "user-events-pipeline",
    "description": "Process and analyze user events from multiple sources",
    "llm_summary": "This workflow extracts user events from Kafka, transforms them with pandas, validates data quality, and loads into data warehouse",
    "steps": [
        {
            "name": "extract-events",
            "description": "Extract user events from Kafka topics",
            "tool": "kafka-consumer",
            "args": {"topics": ["user-clicks", "user-views"]}
        },
        {
            "name": "transform-data",
            "description": "Clean and transform event data using pandas",
            "tool": "python-pandas",
            "depends_on": ["extract-events"]
        },
        {
            "name": "load-warehouse",
            "description": "Load processed data into Snowflake",
            "tool": "snowflake-loader",
            "depends_on": ["transform-data"]
        }
    ]
}

​
3. 🏠 Your Infrastructure Execution

Why this matters: Most AI platforms send your data to their cloud. Kubiya keeps everything in your infrastructure for security, compliance, and performance. 🔒 Security & Compliance Benefits:
  • 🏠 Data Locality: Your data never leaves your infrastructure
  • 🛡️ Zero Trust: Every action validated by your policies
  • 📋 Compliance: GDPR, SOC2 - your rules, your infrastructure
  • 🔐 Air-Gapped: Works completely offline if needed
  • 👁️ Full Visibility: Complete audit trails in your systems
⚡ Performance Benefits:
  • 🚀 Low Latency: Direct access to your resources
  • 🔄 No API Limits: Scale based on your infrastructure
  • 💾 Data Efficiency: No data transfer overhead
  • 🎯 Resource Optimization: Use your existing capacity

​
4. 🧠 LLM-Native Design Philosophy

Everything in Kubiya is designed for AI agents to understand and use effectively: 📝 LLM-Friendly Documentation:
  • Every tool has natural language descriptions
  • Use cases and examples included
  • Common failure modes documented
  • Alternative tool suggestions
🔍 Intelligent Discovery:
  • LLMs can find tools by describing intent
  • Semantic search across all capabilities
  • Auto-suggest related tools and workflows
  • Context-aware recommendations

​
🎯 Next Steps

🚀 Quick Start

Get started in 5 minutes

🔧 MCP Integration

Connect your AI assistant

🛠️ Tool Development

Build your first tools

🏃 Runner Setup

Deploy execution infrastructure
Ready to build enterprise-grade AI applications? Kubiya provides the production infrastructure, security, and scale you need to deploy AI automation that actually works in the real world.