KiwiQ AI Platform is a production-grade, open-source multi-agent orchestration platform designed for building complex AI workflows. It targets developers and enterprises seeking a robust, scalable solution for automating multi-step AI processes, offering benefits like JSON-defined agents, multi-tier memory, and built-in observability. The platform has been battle-tested on over 200 enterprise AI agents.

How It Works

KiwiQ employs an SDK-first approach, defining workflows as Python/JSON graph schemas that compile to LangGraph and are executed via Prefect. This allows for programmatic control over complex logic, routing, and state management. It supports multiple LLM providers (OpenAI, Anthropic, Gemini, etc.) and integrates human-in-the-loop interactions via real-time WebSocket streaming. An event-driven architecture, powered by RabbitMQ, facilitates asynchronous processing between services, while multi-tier memory solutions (PostgreSQL, MongoDB, Weaviate, Redis) manage relational state, versioned documents, vector search, and caching.

Quick Start & Requirements

• Primary Install/Run: Use docker-compose-dev.yml for a full local environment.

  git clone https://github.com/kiwiq-ai/kiwiq-oss.git
  cd kiwiq-oss
  poetry install
  cp .env.sample .env
  # Edit .env with API keys and credentials
  docker compose -f docker-compose-dev.yml up --build
  

• Prerequisites: Python 3.12, Poetry, Docker & Docker Compose. Requires API keys for LLM providers and credentials for databases/message queues.
• Resource Footprint: Running the development environment requires sufficient resources for multiple Docker containers (FastAPI, PostgreSQL, MongoDB, Redis, RabbitMQ, Weaviate, Prefect).
• Links: API Docs: http://localhost:8000/docs, Prefect Dashboard: http://localhost:4201.

Highlighted Details

• Workflow Engine: SDK-first, Python/JSON graph schemas compiled to LangGraph, orchestrated by Prefect.
• Node System: 24+ reusable node types including LLM, routing, data transforms, scraping, code execution, and sub-workflows.
• Pre-built Workflows: 27+ production-ready workflow definitions for content creation, diagnostics, lead scoring, and research.
• Multi-Tier Memory: Integrates PostgreSQL (relational state), MongoDB (versioned documents), Weaviate (vector search), and Redis (caching).
• Observability: Real-time progress via RabbitMQ events, WebSocket streaming, structured logging, and Prefect dashboard integration.
• RAG & Data Management: Built-in RAG pipelines and versioned customer data management via MongoDB.
• Security: Sandboxed code execution and integrated billing/authentication (Stripe, JWT, RBAC).

Maintenance & Community

The project was primarily designed and built by Raunak Bhandari (CTO, KiwiQ AI), with contributions from Gaurav Kumar, Anish Bharadwaj, and Ashish Kumar Behera for specific workflows and frontend development (frontend builder not open-sourced). Contributing guidelines are provided, encouraging forks and pull requests. No specific community channels (Discord/Slack) are listed in the README.

Licensing & Compatibility

The project is licensed under the Apache 2.0 license, which permits commercial use and modification, provided attribution and license text are retained.

Limitations & Caveats

The project includes a disclaimer stating that scraping websites may violate their Terms of Service, and users are solely responsible for compliance. The visual workflow builder mentioned in the description is not part of this open-source release.