HiMarket is officially open source, providing enterprises with a ready-to-use AI open platform
Dao Feng、Shi Mian、Yan Yan
|
Aug 29, 2025
|
HiMarket Officially Open Source
Today, the brand new open source project HiMarket AI Open Platform, jointly initiated by the Higress community and the Nacos community, is officially released.
GitHub Address:https://github.com/higress-group/himarket
As the AI wave sweeps through various industries, we see countless enterprises and developers enthusiastically participating. Whether it is enabling internal large model capabilities for various business lines, or hoping to open up expensive AI assets and build an ecosystem, a powerful "AI Open Platform" has become a must-have.
As contributors of Higress, we have found in our interactions with many developers and enterprise users that when enterprises use the Higress AI Gateway to complete unified access to models and tools, the challenge shifts from "connectivity" to "collaboration and productization." Various roles such as tool providers, Agent developers, security compliance, and financial operations are operating in silos, leading to issues such as scattered capability catalogs, inconsistent listing processes, difficulties in unifying permissions and quotas, challenges in measuring and allocating costs, and ensuring content security in every call. Enterprises naturally need a multi-role oriented, user-friendly, and functionally complete upper-level open platform. So, how should enterprises build such a platform?
Starting from scratch, you might need:
Develop a complete portal system (> 3 months): from UI/UX design to front-end and back-end development.
Implement complex developer and application management: registration, approval, RBAC permissions, credential management and rotation.
Design complex subscription and authorization processes: How to allow developers to subscribe independently? How to securely sync authorization relationships to the gateway?
Build observability for operations and developers: pull raw data from the gateway and MCP Registry for secondary development, and perform multi-dimensional cost and measurement analysis by model and by consumer.
We estimate that this part of the work accounts for 80% of the effort to build a complete AI open platform. This not only takes time and effort but may also hinder the progress of AI deployment in enterprises.
This is exactly the original intention of our initiative HiMarket: to help users start building the AI open platform from 80%.
What is HiMarket?
HiMarket is a ready-to-use AI open platform solution. It is not just a simple API listing page, but an enterprise-level “AI Capability Market and Developer Ecosystem Center”.
It consists of several core components, perfectly matching the needs of different roles within the enterprise:
AI Open Platform Management Backend (for Administrators/Operations): Here, the underlying model services, MCP Server, Agents, and other diverse AI capabilities can be easily packaged as standardized “AI products” in API form, complete with comprehensive documentation and examples, and finally published to the portal with one click.
AI Open Platform Portal (for Developers): The portal acts as a “store” for internal and external developers. Developers can complete developer registration, create consumers, obtain credentials, browse and subscribe to AI products, conduct online testing, and clearly monitor their calling status and costs.
AI Gateway: As a sub-project of the Higress community, the Higress AI Gateway handles all AI call authentication, security, traffic control, protocol conversion, and observability capabilities.
Nacos: Nacos serves as the MCP Registry, providing comprehensive MCP Server metadata hosting, version management, service discovery, key management, and more for the portal. It achieves flexible expansion and changes in the system through dynamic service discovery and dynamic configuration changes and simplifies the management and invocation of multiple MCP services through the Nacos MCP Router smart routing. In the daily development testing scenario, it realizes efficient MCP service debugging through dynamic prompt changes and implements grayscale release through multi-version management in production environments, enhancing security capabilities through encrypted storage and dynamic configuration changes.
With HiMarket, you can focus on the core value creation:
Access your core AI capabilities: whether it’s self-developed models or third-party services.
Define your AI products: package AI capabilities into commercially viable AI products.
Operate your developer ecosystem: invite developers to join, support enterprise-level internal user systems, and build partnerships.
HiMarket can provide assistance for the following scenarios
Central AI Platform within Enterprises
Within large enterprises, when various business lines begin to introduce AI capabilities, if each operates independently, it will quickly fall into chaos in access, cost control, and security risks, affecting the overall speed of AI innovation in the enterprise. At this time, HiMarket, as a unified AI asset management platform in the AI era, becomes a unified and standardized entry point for all internal systems to access external AI services. This centralized model brings great management benefits:
Cost Auditing and Allocation: Accurately record and analyze the Token consumption of various business departments and applications, providing clear and credible data support for cost control, budget allocation, and internal settlements.
Data Security and Compliance: Establish a unified content security policy for the entire enterprise, effectively preventing sensitive internal data (such as customer information and financial reports) from being inadvertently or maliciously sent to external models through prompts, thereby building a crucial data leakage prevention barrier.
Resource Reuse: By deploying a unified semantic cache, return cached results directly for high-frequency, similar questions across the entire company, effectively reducing model invocation costs and significantly improving response speed, enhancing employee and customer experience.
Efficiency Improvement: Use a unified AI platform to connect departmental data, reuse private domain data retrieval/Q&A/diagnostic capabilities, standardize access and delivery processes, and significantly accelerate the deployment of AI capabilities across various business lines.
Private MCP Market: A Unified Entry Point for the Enterprise MCP Tool Ecosystem
With the popularization of the MCP protocol, enterprises can connect models to external tools and services in a “plug-and-play” manner through standardized interfaces, greatly reducing integration costs and enhancing system flexibility and scalability. Internally, enterprises centralize access and governance of MCP services through the MCP market as a single entry point, enabling AI Agents to efficiently execute specific tasks such as order inquiries and conference room bookings using a unified manner while ensuring consistency in traffic governance and elastic scaling. All outbound tool calls from the enterprise must first go through HiMarket for centralized, standardized security control, including unified authentication and authorization, fine-grained rate limits, comprehensive audit logs, etc. This method avoids the repetitive implementation and maintenance of complex security logic on every independent tool service, greatly simplifying the security governance of the MCP tool ecosystem, providing a solid foundation for enterprises to build a secure, reliable, and controllable AI Agent system. HiMarket can help enterprises quickly launch a privatized MCP market, such as the open-source MCP market built by Higress: https://mcp.higress.ai/
Build a Unified AI Open Platform for the Monetization of Enterprise AI Capabilities
The ultimate goal of any enterprise is revenue and profit. As AI agents accelerate their entry into work and life, more and more voices in the industry are calling this trend the “Agent Economy.” From the imagination of the ecological map, it can be seen that tools, data, computing power, and industry services around intelligent agents will be standardized into tradable capabilities, offering broad commercial opportunities. The core carrier that supports these monetization paths is the “AI Open Platform” aimed at developers and enterprises—connecting the perception, reasoning, and external tools of intelligent agents through a unified API to achieve large-scale distribution and billing.
HiMarket aims to become the open-source foundation of the “AI API Economy,” helping enterprises unify distributed intelligent agents, models, RAG, tools, and industry data into standardized API assets. Through a one-stop capability catalog and listing, access verification, permissions and auditing, invocation measurement and quotas, as well as billing and reconciliation interface docking, HiMarket enables capabilities to be quickly discovered, combined, distributed, and settled, forming a closed loop from internal capabilities to external revenue, and creating a unified AI open platform and API monetization channel that belongs to enterprises.
HiMarket: An Open Platform Powered by AI
In line with the evolution towards the AI era, HiMarket's vision is to become an “AI-powered open platform”: leveraging an understanding of OpenAPI, MCP Manifest, OAS Swagger, and other API metadata to explore AI-assisted one-click generation of multilingual SDKs, examples, and contract testing; achieving intelligent discovery and routing of tools based on semantic retrieval and task intent; forming quality profiles and scores for availability, latency, success rate, etc., during runtime; providing Token cost measurement and budget optimization; and implementing content security at the prompt level to prevent sensitive information leakage.
The above capabilities are currently in the roadmap stage and have not been fully realized. We hope to verify and refine them together with the community—promoting prototype implementation through open RFCs and a modular architecture, prioritizing the seamless integration of the “metadata ingestion—assessment benchmarks and telemetry standards—quality scoring and routing strategies” minimum viable closed loop. We welcome developers to participate in discussions, submit issues/PRs, or contribute adapters and assessment sets, to make HiMarket a unified entry point and innovation accelerator for the enterprise private AI API market in “discovery—access—governance—monitoring—optimization—monetization.”
Quick Start
We have prepared a detailed quick start guide for you. With just a few simple steps, you can run HiMarket locally and personally experience the thrill of “starting from 80%.”
Step 1: Prepare Environment and Code
Ensure you have installed Git, JDK 8, Maven, Node.js (v20), and have a MySQL database ready.
Step 2: Start Backend Service
The HiMarket backend service relies on the database. Please modify the configuration according to portal-bootstrap/src/main/resources/application.yaml, or pass it in as a parameter during startup.
Step 3: Start the Frontend
HiMarket includes both the management backend and the developer portal frontends.
Step 4: 5-Minute Experience of Core Processes
Access the Management Backend localhost:5174, register an administrator account.
Import Service Source
Select the service source where the MCP Server is located according to the business scenario, currently supporting both the gateway and Nacos.
a. Import Gateway Instance: In [Instance Management], import the Higress gateway.
b. Import Nacos Instance: In [Instance Management], import Nacos.
Create Portal: In the [Portal] menu, create your own AI portal and bind it to a domain name for local testing (such as localhost).
4. Create and Publish AI Products: In [API Products], package a service from Higress's MCP into a product and publish it to the portal you just created.
Switch to the Developer Portal (localhost:5173), register a developer account, and wait for the administrator’s approval in the backend.
Subscribe and Call: Log into the developer portal, and you will see the AI product published earlier. Create a consumer to obtain credentials, then subscribe to the product to start calling!
Our Journey and Your Participation
HiMarket is currently in its initial stage. We have implemented the core processes of the AI open platform, but this is just the beginning.
HiMarket Roadmap:
Phase 1 (Current): Release a fully functional open-source MVP version, polish with the community, and optimize the experience.
Phase 2 (Q3 2025): Observability, measurement billing, splitting REST API Model API/Agent API, A2A support.
Phase 3 (Q4 2025): MCP quality tools, RBAC, portal DIY, debugging, SDK generation.
The future of HiMarket relies on the participation and contributions of every developer. We sincerely invite you to:
Star Our Project: This is the most direct encouragement to us https://github.com/higress-group/himarket
Deploy and Try: Use HiMarket and experience its value.
Submit Issues: Whether it’s a bug or a feature suggestion, your feedback is crucial.
Contribute Code: We have prepared a series of good first issues and look forward to your first PR.
Join the Community: Scan the QR code below to join our community discussion group (DingTalk), and interact directly with core developers.
Co-build the Ecosystem: Maintain a code structure that is friendly for secondary development and expansion, welcome to fork customization and contribute back, and work together to improve the ecosystem.
