KDCube provides an open‑source, self‑hosted production runtime that lets developers deploy AI agents and applications such as LangGraph, CrewAI, or Claude Agent SDK without rewriting their logic. The platform supports multi‑user chat, streaming, user‑scoped state, spend controls, and isolated code execution, with built‑in telemetry and Git‑based delivery for reliable, cost‑tracked deployments. It enables teams to run AI workloads on their own infrastructure while maintaining full control over security, licensing, and scaling.
Funding
Funding not disclosed
Founders
Product
Problem
KDCube addresses the difficulty of deploying and managing AI agent applications—such as LangGraph, CrewAI, or Claude Agent SDK—in production environments. Teams often need to handle multi‑user interactions, state isolation, spend control, and secure code execution while maintaining a single, maintainable deployment pipeline.
Solution
KDCube provides an open‑source, self‑hosted runtime that centralizes the execution of AI agents across many users and applications. The platform isolates code execution per tenant and user, enforces spend limits before work begins, and tracks telemetry for each interaction. Integration with Git enables continuous delivery of agent logic without altering existing codebases. Developers can expose agents through chat widgets, REST APIs, webhooks, or real‑time streams, while the runtime manages authentication, routing, and resource governance. The system supports both built‑in agents and external stacks, allowing teams to keep their preferred agent frameworks while adding production‑grade features such as budgeting, state management, and secure tool execution.
Target Audience
Primary customers are development teams and product engineers building AI‑driven applications that require scalable, secure, and cost‑controlled production runtimes, including SaaS providers, enterprise AI platforms, and internal tool developers.
Features
- Multi‑tenant architecture with user‑scoped workspaces and isolated code execution
- Built‑in spend controls that attribute costs per user, app, and turn, with budget checks before execution
- Unified deployment pipeline from Git, supporting continuous delivery of agent updates
- Flexible API surfaces including chat widgets, REST/webhooks, SSE, Socket.IO, and channel integrations (e.g., Telegram, Slack)
- Integrated telemetry and execution logs for monitoring, debugging, and performance analysis
- Edge governance layer providing TLS routing, authentication, admission control, and file scanning
- Support for both native KDCube agents and external frameworks like LangGraph, CrewAI, and Claude Agent SDK