What is MxPage?

MxPage is an open-source AI e-commerce image workspace maintained by 灵矩绘境 (MatrixInspire). It helps e-commerce operators, cross-border sellers, designers, and self-hosted teams create product hero images, complete detail page long images, and Xiaohongshu vertical posts. Users upload reference images and product facts, select platform presets (generic, Taobao, Pinduoduo, Douyin, Xiaohongshu) and language, then the V4 deterministic workflow handles product recognition, value proposition planning, storyboarding, and image generation with a prompt confirmation step before generation. It also supports batch processing of multiple products by folder, image translation with fixed prompts (no re-planning), and an Agent workspace powered by LangGraph.js for conversational task execution. The software runs as a local Electron desktop app on Windows x64 and macOS (Apple Silicon and Intel), or from source with Node.js 24 and npm using an embedded SQLite database via Prisma. All AI model calls use the user's own OpenAI-compatible API endpoints and keys; costs are paid directly to the model provider. The codebase is MIT licensed.

Should you self-host MxPage?

MxPage scores 48/100 on selfhost.directory (how scoring works) and ranks #5 of 10 in E-commerce. It has 445 GitHub stars.

It shipped 8 stable releases in the past year. The latest, v0.3.3, came out on 20 October 2026. New versions land roughly every 2 weeks.

Setup is moderate: expect some configuration beyond a single command. It is MIT-licensed and written mostly in Typescript.

Its closest alternative on selfhost.directory is BTCPay Server (76/100, 7 860 stars). It scores higher overall, so compare the two before you pick. See MxPage vs BTCPay Server side by side.

Features

Agent workspaceDesktop control drivers integrated but default off; Windows/macOS GUI acceptance not verified
AI product image generation
Batch Processing
Cross-platform desktop appsNo commercial code signing or Apple notarization; macOS uses ad-hoc signing
Data backup and restore
Detail page generation
Deterministic V4 workflow
Image editing with history
Image translation
Local Data Storage
Model discovery and selection
Monitoring and cost tracking
Multi-platform visual presets
OpenAI-compatible API support
Xiaohongshu post generation

FAQ

Can it auto-publish to Taobao, Xiaohongshu, or control my computer?

Platform presets are for generating images and copy only, not auto-publishing to merchant accounts. Agent desktop control is default off; Windows/macOS real GUI acceptance status is documented separately.

Does local deployment mean images never leave my computer?

No. Works and database can stay local, but images and descriptions for generation are still sent to your configured model service. Keys are stored in the current browser and provided to the local service at request time.

Will disconnection or timeout cause duplicate charges?

Completed steps are reused via checkpoints. Saved upstream task IDs continue querying the original task; uncertain requests are not auto-retried and require explicit handling. Model fees are ultimately per provider billing.

Does image translation re-plan the entire detail page?

No. Translation uses an independent fixed-prompt workflow providing original and translated image comparison. Product editing, translation, and export results should still be manually verified.

Which model services are supported? Do I need to manually select protocols?

Enter your service address and API key, discover models and save per-role defaults. Text/vision auto-compatible with Responses, Chat Completions, Anthropic Messages, and Gemini; images support standard Images, Gemini, and described EvoLink async interfaces.

Can I generate only product hero images or only detail images?

Yes. Detail tasks support 0-10 hero images and 0-20 detail images, with at least one total. Xiaohongshu independently supports 1-8 vertical images, defaulting to five.

Is MxPage a free AI e-commerce detail page generator?

The source code is MIT licensed and can be self-deployed. AI model calls are not free - product recognition, planning, image generation, editing, and translation may incur charges from your configured provider. Test with one image first to verify model, cost, and effect.