loopstack.ai

What It Is

Loopstack is an AI workflow framework for TypeScript and NestJS, designed for building stateful automation, AI Agents, and interactive human-in-the-loop processes. It is not an end-user chat tool, but a backend development framework. Its core components are Workflows, Tools, and Documents, with NestJS modules extending capabilities such as LLM Providers, OAuth, sandboxing, and Secrets.

Core Capabilities

Its main value lies in bringing “deterministic workflows” and “agentic LLM calls” into a single system. Workflows are defined as TypeScript state machines and support transitions, guards, wait transitions, dynamic routing, sub-workflow callbacks, and state persistence. On the AI side, it supports text generation, structured output, tool calling, multi-turn chat, Agent loops, switching between multiple Providers, streaming output, and Prompt templates. The documentation mentions Claude and OpenAI, and custom LLM Providers can also be implemented. Human-in-the-Loop is a first-class framework primitive, allowing workflows to pause for hours or days while waiting for user confirmation, form submission, or approval.

Pricing and Deployment

The documentation clearly states that Loopstack is free and open source under the MIT License. No hosted version, enterprise edition, or commercial pricing was found. For deployment, it requires Node.js, NestJS, Docker, and depends on PostgreSQL and Redis. React Studio can be used to run, debug, and organize automations. LLM calls themselves require external model API keys, so actual costs depend on third-party services such as Claude and OpenAI.

Pros and Cons

Its strengths lie in strong engineering control: state checkpoints, automatic retries, timeouts, error states, manual retries, and audit trails are all part of the design, making it suitable for production-grade Agents. It also avoids handing the entire business process over to the model, allowing deterministic logic and human confirmation to be added at critical points. The downside is that the learning curve is clearly higher than low-code tools. Users need familiarity with TypeScript, NestJS, Docker, and backend deployment. The documentation does not show Chinese-language support, commercial SLAs, or hosted service information.

Who It’s For and Access from China

Loopstack is better suited to teams that already use a NestJS/TypeScript stack, for scenarios such as customer support tickets, code review, approval workflows, internal automation, and complex Agent orchestration. For users who simply want to quickly build no-code AI applications, Dify, Flowise, or n8n may be more straightforward. Access from mainland China is not clarified in the available text. Domain availability, GitHub/npm downloads, and access to and payment for Claude/OpenAI APIs may all be affected by network and account conditions, so real-world testing is required. Domestic alternatives worth considering include Dify, FastGPT, and self-hosted n8n.