kdcube
Active·★ 26·MIT·Updated 2026-10-21
KDCube is an open-source, self-hosted production runtime designed for AI applications and agents, providing the necessary infrastructure to run frameworks like LangGraph, CrewAI, or custom agents in production. It handles multi-user serving, ordered conversation delivery, streaming, isolated code execution, cost controls, and secure deployment from Git, allowing developers to focus on agent logic rather than operational complexities.
#AI Runtime#Agent Framework#Self-Hosted#Multi-User Serving#Deployment#Isolated Execution#Cost Control#API Integration
01
Features
01Multi-user serving with ordered conversation delivery and streaming capabilities.
02Secure deployment and updates of app code from Git, with descriptor-based configuration and near-live app reloads.
03Runs generated code under explicit isolation policies, including split-Docker profiles for enhanced security.
04Built-in user identity, connected accounts, delegated operators, and grants for secure access management.
05Tracks economics for LLM, embedding, web-search, and tool usage, enforcing budgets before calls.
02
Compatibility
Python
Runtime
Verified via docs
Docker
Deployment
Verified via docs
LangGraph
Agent Framework
Verified via docs
CrewAI
Agent Framework
Verified via docs
Claude Agent SDK
Agent Framework
Verified via docs
03
Quick start
1
$ pip install kdcube-cli
04
Use cases
↳Adding multi-user delivery, streaming, persistence, and deployment to existing LangGraph, CrewAI, or custom agents.
↳Integrating a configurable chat widget or native APIs into a website or product UI.
↳Managing and tracking tools, provider integrations, user consent, and isolated execution for AI services.
↳Quickly building new AI features with ready chat, ReAct agents, web search, files, and user memory.
↳Orchestrating several independently deployable AI services or frontends that provide and consume APIs.
05
Alternatives
lobu★ 230
Open-source control plane and runtime for organisational agents: shared company context, isolated execution, approvals and MCP.
toolport★ 235
Local-first MCP gateway. One port for every tool and every AI client: lazy discovery (~90% token savings), tool integrity + quarantine, secrets in the OS keychain.
tachyon★ 37
💫 Tachyon MCP Runtime for Java and Kotlin - Model Context Protocol (MCP) server with Streamable HTTP, native Netty transports, tools, tasks, resources, prompts, completions, resumable sessions, and stateless deployment support.
oto-dock★ 193
The self-hosted platform for your company's AI agents. Your team builds them on Claude Code, Codex or local models, and they work in chats, on schedules, on the phone and on your own machines.
Equibles★ 237
An open-source, self-hosted mini Bloomberg Terminal for AI agents, exposed as an MCP server — SEC filings, institutional holdings, insider trading, congressional trades, and short data from a single self-hosted stack.
Comments
Log in to leave a comment
No comments yet. Be the first!