Platform + SDK for Agentic Chat Applications

Self-hosted platform and SDK for building and operating agentic chat applications and copilots. Production-ready runtime with streaming, tool execution, memory, economics, and enterprise governance—all out of the box. Deploy in 30 minutes.

Framework-agnostic runtime — Build with LangGraph, LangChain, CrewAI, AutoGen, or custom Python
Multi-tenant production hosting with schema-level isolation and audit trails
Sandboxed code execution with ephemeral Docker containers and resource limits
Built-in economics system — 60-80% cost savings vs. managed platforms
Self-hosted data sovereignty
Zero vendor lock-in (MIT license)
Production knowledge base with pgvector and hybrid search
Dynamic bundle system for agent hot-reload
Star on GitHub LinkedIn Support MIT License

Framework-Agnostic Runtime

Build your workflows in LangGraph, LangChain, CrewAI, AutoGen, or custom Python. KDCube provides the runtime infrastructure to host, scale, and govern them with streaming, tool execution, and memory management out of the box. Runtime architecture →

Governance & Compliance

Enterprise Security & Governance

Production-ready multi-tenant isolation with per-tenant/project schemas, comprehensive audit trails, and self-hosted data sovereignty. Built-in authentication, RBAC, sandboxed execution, and complete compliance controls for SOC2/HIPAA deployments. Security & governance →

Complete Runtime Stack

Streaming chat (REST/SSE/Socket.IO), tool execution, code sandboxing, knowledge base with pgvector, web research, memory management, and artifact handling. Everything needed to run production agentic applications. Key features →

Economics & Cost Management

Built-in usage tracking, budget enforcement, rate limiting, and cost reporting at tenant/project/user levels. Real-time dashboards and token-level granularity. Achieve 60-80% cost reduction vs. managed platforms with zero SaaS fees. Economics system →

Reference Agent Implementations

Example agents demonstrating runtime capabilities—fork and customize

Knowledge Agent

Enterprise RAG with pgvector, hybrid search, citations, and multi-tenant isolation.

Code Execution Agent

Sandboxed Python execution in ephemeral Docker containers with resource limits.

Research Agent

Multi-backend web search with Brave and DuckDuckGo integration.

Strategic Reasoning Agent

ReAct agent with planning, tool selection, and adaptive reasoning capabilities.

Document Processing Agent

Multi-format document processing with extraction and analysis capabilities.

Marketing Writer Agent

Marketing content generation with brand compliance and SEO optimization.

View Complete Agent Definitions →

Why Teams Choose KDCube

The enterprise-grade application server for AI agents

vs. Managed Platforms

Vellum, StackAI, Vertex AI

  • 60-80% cost reduction — No SaaS fees or per-pipeline markups
  • Full control — Your infrastructure, your scaling rules
  • Data sovereignty — Pass audits on your terms, nothing leaves your VPC
  • Zero lock-in — Switch providers in hours with MIT license

vs. Agent Frameworks

LangChain, CrewAI, AutoGen

  • Complementary layer — Build with frameworks, deploy on our platform
  • Production infrastructure — Streaming runtime, tool execution, memory, economics
  • Multi-framework support — Run LangGraph, LangChain, CrewAI on one platform
  • Enterprise features — Multi-tenancy, knowledge base, sandboxing, cost tracking

vs. Custom Build

DIY Infrastructure

  • Deploy in 30 minutes — vs. 3-6 months of custom development
  • Battle-tested — Streaming, tool execution, memory, economics, governance solved
  • Focus on agents — Build application logic, not runtime infrastructure
  • Full source access — MIT licensed, fork freely

Enterprise-Grade Agent Deployments

Mission-critical deployments from regulated industries to high-growth startups

Regulated Industries

Deploy agentic applications with SOC2/HIPAA compliance controls, schema-level tenant isolation, comprehensive audit trails, and self-hosted data sovereignty. Pass audits without changing agent code.

30% faster compliance certification (Forrester)

Platform Engineering Teams

Provide production-ready runtime infrastructure for teams building agentic applications. Multi-tenant platform with streaming, tool execution, knowledge base, and economics—the scalable foundation your teams need.

$150K-$500K/year SaaS costs eliminated (IDC)

AI-Powered Products

Ship AI-powered products with production-ready runtime and cost control. Build agents with any framework, KDCube handles streaming, tools, memory, and governance infrastructure.

70-90% cost predictability improvement

AI Startups

Launch agentic products with production runtime out-of-the-box. Multi-tenancy, streaming, tools, knowledge base, and economics included. Focus on your agent logic, not infrastructure.

70-90% faster prototype-to-production

Frequently Asked Questions

How does the application server work with different frameworks?

KDCube is framework-agnostic—build your workflows in LangGraph, LangChain, CrewAI, AutoGen, or custom Python. The platform provides the runtime infrastructure (streaming, tools, memory, economics, governance) to host and scale them. Architecture overview →

What governance features support compliance?

Schema-level multi-tenant isolation, comprehensive audit trails, SOC2/HIPAA controls, delegated authentication with 2FA, RBAC, encryption, and complete execution lineage. Self-hosted data sovereignty. Security & governance →

How does sandboxing work?

Ephemeral Docker containers with network isolation, privilege separation, and resource limits (CPU/memory/time). Tool calls proxied through supervisor to prevent data exfiltration. Sandboxing details →

How long does deployment take?

20-35 minutes via Docker Compose or CLI installer. Configure database, Redis, LLM keys, and start services. Custom-built platforms require 3-6 months and $150K-400K.

What expertise is required?

Basic Docker Compose and database connectivity. No specialized MLOps or Kubernetes expertise required for getting started.

How does cost tracking work?

Automatic per-tenant, per-project, per-user tracking with token-level granularity. Real-time dashboards, budget enforcement, rate limits, and comprehensive reporting built-in. Economics system →

Can I migrate existing agents?

Yes. LangGraph workflows run with zero code changes. LangChain requires minimal adapters. Custom Python agents integrate via the Bundle API.

What's included out-of-box?

Included: Streaming runtime, tool execution, code sandboxing, knowledge base, memory, economics, auth, multi-tenancy, monitoring. Configure: LLM keys, database, Redis, your agent bundles.

Is this enterprise production-ready?

Yes. Built-in circuit breakers, rate limiting, backpressure control, retries, comprehensive audit logging, and horizontal scaling. Production deployments handle 10,000+ executions/day.

What are the ongoing costs?

Infrastructure only: compute ($200-2K/mo), storage ($50-500/mo), LLM API costs. No per-pipeline, per-tenant, or SaaS fees. Platform is free and open source (MIT license).

How do compliance audits work?

Self-hosted means you control the security posture. Platform provides schema-level isolation, comprehensive audit trails, RBAC, encryption, and execution lineage. Pass audits 30% faster. Compliance controls →

What support is available?

Community: GitHub issues, docs (free). Enterprise: SLA support, consulting, custom development.

Documentation & Resources

Deployment Guides

K3s Quick Start

Single-node K3s deployment with SSL and monitoring.

K3s Production

Production deployment with HA and backup strategies.

Kubernetes Local

Full Kubernetes for development and staging.

AWS Deployment

AWS infrastructure with ECS, RDS, and ElastiCache.

Deployment Comparison

Compare AWS, K3s, and on-premises with cost analysis.

Chat Widget Integration

Website chat widget with WebSocket integration.