Kailash SDK Documentation
Enterprise AI Agent Platform with Cryptographic Trust
The Kailash SDK is the enterprise-grade platform for building production AI agent workflows with cryptographic trust chains, multi-channel deployment, and zero-config database operations. From a single workflow to orchestrated multi-agent systems, Kailash provides the foundation for verifiable, auditable AI applications.
Current Release: v2.59.0 (Core SDK) | DataFlow v2.19.0 | Nexus v2.14.0 | Kaizen v2.41.0
Quick Links
- Getting Started
New to Kailash? Start with Installation and your first workflow.
- Core SDK
Build workflows with WorkflowBuilder, execute with LocalRuntime or AsyncLocalRuntime, and secure with CARE trust chains.
- Frameworks
Kaizen (v2.41.0): AI agents with signatures, multi-agent coordination, and CARE trust
Nexus (v2.14.0): Multi-channel platform – API + CLI + MCP from one codebase
DataFlow (v2.19.0): Zero-config database with 11 auto-generated nodes per model, including FieldType.Vector(dim) embedding columns
- Enterprise
Security, compliance, monitoring, and deployment.
Getting Started
Enterprise
Examples & Patterns
API Reference
Development
Architecture Overview
The Kailash SDK provides a layered architecture:
+----------------------------------------------+
| Frameworks (Choose One+) |
| Kaizen (AI Agents) | Nexus (Multi-Channel) |
| DataFlow (Database) | MCP (Integration) |
+----------------------------------------------+
| Core SDK Foundation |
| WorkflowBuilder | Runtime | Nodes | Trust |
+----------------------------------------------+
| User Infrastructure (not part of SDK) |
| Docker | Kubernetes | Cloud | Monitoring |
+----------------------------------------------+
Critical relationships:
DataFlow, Nexus, and Kaizen are built ON the Core SDK – they do not replace it
All frameworks share the same workflow execution:
runtime.execute(workflow.build())CARE trust is woven through every layer for cryptographic accountability
What Makes Kailash Different
- Cryptographic Trust (CARE/EATP)
Every agent action traces back to a human authorization through verifiable delegation chains. No other Python AI framework provides built-in cryptographic trust at the runtime level. See CARE Trust Framework.
- Multi-Channel from One Codebase
Write a workflow once. Deploy it as a REST API, CLI tool, and MCP server simultaneously with Nexus. See Nexus – Multi-Channel Platform.
- Zero-Config Database Operations
Declare a model with
@db.modeland get 11 production-ready database nodes automatically – CRUD, bulk operations, counting, and upserting. See DataFlow – Database Framework.- Signature-Based AI Agents
Define agent behavior with signatures, not prompts. Agents automatically optimize, fall back to alternative models, and coordinate in multi-agent teams. See Kaizen – AI Agent Framework.