HiMarket is an enterprise-grade AI open platform built on Higress AI Gateway, helping enterprises build private AI capability marketplace to uniformly manage and distribute AI resources such as LLM, MCP Server, Agent, and Agent Skill. The platform encapsulates distributed AI capabilities into standardized API products, supports multi-version management and gray-scale release, includes a built-in Skills Marketplace for developers to browse and install Agent Skills, provides HiChat AI conversation and HiCoding online programming for self-service developer experience, and features comprehensive enterprise-level operation capabilities including security control, observability analysis, metering and billing, making AI resource sharing and reuse efficient and convenient.
| Category | Feature | Description |
|---|---|---|
| AI Marketplace | Model Marketplace | Integrate various models with content safety, token rate limiting, and other protection capabilities |
| MCP Marketplace | Integrate MCP Servers from various platforms, support converting external APIs to standard MCP Servers | |
| Agent Marketplace | Package and publish Agent applications, integrate with AgentScope and other Agent building platforms | |
| Skills Marketplace | Upload and distribute Agent Skills, developers can browse, subscribe, and install Skill packages | |
| AI Experience Center | HiChat Conversation | Single-model conversation and multi-model comparison, MCP tool invocation testing, enhanced features like web-connected Q&A |
| HiCoding Online Programming | Integrated secure sandbox environment, supporting Vibe Coding and human-AI collaborative development with real-time file changes and code preview | |
| Enterprise Management | Product Management | Authentication, traffic control, call quotas, and other protection capabilities |
| Observability | Full-chain monitoring, call tracing, heatmaps, anomaly alerts | |
| Metering & Billing | Token-based call counting with automatic cost statistics | |
| Version Management | Multi-version parallel operation, gray-scale release, quick rollback | |
| Customization | Portal Branding | Custom domain, logo, color scheme, page layout |
| Identity Authentication | Support third-party OIDC integration with enterprise identity systems | |
| Approval Workflow | Configurable auto/manual approval for subscription and product scenarios | |
| Product Catalog | Custom category tags with browsing, filtering, and search support |
System Architecture
HiMarket system architecture consists of three layers:
- Infrastructure: Composed of AI Gateway, API Gateway, Higress and Nacos. HiMarket abstracts and encapsulates underlying AI resources based on these components to form standard API products for external use.
- AI Open Platform Admin: Management platform for administrators to create and customize portals, manage AI resources such as MCP Server, Model, Agent, and Agent Skill, including setting authentication policies and subscription approval workflows. The admin portal also provides observability dashboard to help administrators monitor AI resource usage and operational status in real-time.
- AI Open Platform Portal: Developer-facing portal site, also known as AI Marketplace or AI Hub, providing one-stop self-service where developers can complete identity registration, credential application, product browsing and subscription, online debugging, and more. Developers can also interact with models and MCP Servers through HiChat, or perform online AI programming in secure sandboxes through HiCoding.
Admin Dashboard |
Developer Portal |
Option 1: Local Setup
Requirements: JDK 17, Node.js 18+, Maven 3.6+, MySQL 8.0+
Start Backend:
# Build project
mvn clean package -DskipTests
# Start backend service
java --add-opens java.base/java.util=ALL-UNNAMED \
--add-opens java.base/java.lang=ALL-UNNAMED \
--add-opens java.base/java.lang.reflect=ALL-UNNAMED \
-Ddb.host=${DB_HOST} \
-Ddb.port=${DB_PORT} \
-Ddb.name=${DB_NAME} \
-Ddb.username=${DB_USERNAME} \
-Ddb.password=${DB_PASSWORD} \
-jar himarket-bootstrap/target/himarket-bootstrap-1.0-SNAPSHOT.jar
# Backend API: http://localhost:8080Start Frontend:
# Start admin portal
cd himarket-web/himarket-admin
npm install
npm run dev
# Admin portal: http://localhost:5174
# Start developer portal
cd himarket-web/himarket-frontend
npm install
npm run dev
# Developer portal: http://localhost:5173Option 2: Docker Compose
Requirements: Docker, Docker Compose
Script Deployment: Use the interactive install.sh script to deploy the full stack (HiMarket, Higress, Nacos, MySQL) with guided configuration.
git clone https://github.com/higress-group/himarket.git
cd himarket/deploy/docker
./install.shAI Deployment (Recommended): If you're concerned about environment compatibility issues during deployment, we recommend using AI Coding tools such as Cursor, Qoder, or Claude Code, which can automatically detect and resolve environment problems. After cloning the project, simply enter in your AI tool:
Read the deployment docs under deploy/docker and help me deploy HiMarket with Docker Compose
See the Deployment Guide for details.
Service URLs after deployment:
- Admin Portal: http://localhost:5174
- Developer Portal: http://localhost:5173
- Backend API: http://localhost:8081
Uninstall:
./install.sh --uninstallOption 3: Helm Chart
Requirements: kubectl (connected to a K8s cluster), Helm
Script Deployment: Use the interactive install.sh script to deploy HiMarket to a Kubernetes cluster with guided configuration.
git clone https://github.com/higress-group/himarket.git
cd himarket/deploy/helm
./install.shAI Deployment (Recommended): If you're concerned about environment compatibility issues during deployment, we recommend using AI Coding tools such as Cursor, Qoder, or Claude Code, which can automatically detect and resolve environment problems. After cloning the project, simply enter in your AI tool:
Read the deployment docs under the deploy directory and help me deploy HiMarket to my K8s cluster with Helm
See the Deployment Guide for details.
Uninstall:
./install.sh --uninstallOption 4: Cloud Deployment (Alibaba Cloud)
Alibaba Cloud ComputeNest supports out-of-the-box deployment of the community edition with one click:
For detailed usage instructions, please refer to:
DingTalk Group |
WeChat Official Account |
Thanks to all the developers who have contributed to HiMarket!