
NavXD Review 2026
An AI tool radar that sorts new tools by the job you want done
Dozens of AI tools launch every day, and NavXD's answer is to treat them like a radar screen rather than a list. The site scans 1,288 signals from 11 sources, logs each new tool with a category, a signal strength and a timestamp, and asks one question first: what are you trying to do?
The current version, AI Tool Radar v3, ranks 118 tools across 35 categories, runs 86 in-depth reviews and 65 scene workflows, and switches between Chinese and English from the header. Behind it is a live feed that in a single day picked up a new GLM model release, a Qwen open-source drop and the OpenAI Agents SDK, CrewAI, AutoGen and LangGraph entering the index.
Organised by task, not by vendor
Five base scenarios hold every tool, route, comparison and ranking on the site.
AI Coding
Development, site building, SaaS and code refactoring. Linked tasks: build a website, build a SaaS, fix a bug.
Content Generation
Text, images, video, audio and marketing content. Linked tasks: short video, Xiaohongshu posts, ad creatives.
Knowledge / RAG
Document Q&A, PDFs, knowledge bases and research. Linked tasks: paper summaries, company knowledge bases, competitor analysis.
AI Agent
Breaking tasks down, executing them and multi-agent collaboration. Linked tasks: automated research, automated reports, agent apps.
Automation / Workflow
Chaining tools, scheduled jobs and automatic publishing. Linked tasks: weekly reports, auto-posting, competitor scraping.
On top of the scenarios sit recommended routes that chain single tools into a working path. The content route goes from topic research in Kimi to writing, illustration in Jimeng and publishing through Feishu AI; the coding route runs requirements through DeepSeek, generation in Trae, debugging in Tongyi Lingma and deployment in Coze. Each of the 65 scene workflows comes with steps, a tool combination, prompt templates and a note on how to make money with it.
Rankings, reviews and trend reports
The leaderboard scores tools on capability, speed, price, experience and availability from mainland China, and is recalibrated every week. Each ranked tool has a long-term tracking record: a hands-on review with real task scenarios, a comparison with its peers, and notes on price, limits and alternatives.
Monthly trend reports cover how categories evolve, which players move, how prices curve and where the technical milestones land. Twelve capability channels, from large language models and coding to image, video, audio, writing, design, office, marketing and SEO, education, lifestyle and agents, gather every tool, comparison and tutorial in one field.
Anyone can submit a tool, and submissions are reviewed within 48 hours. An affiliate marketplace collects partner deals such as cashback, trial credit and education discounts, and a data API and a monthly newsletter round out the offer.
What Pro adds
Pro costs ¥99 a year or ¥12 a month, cancel any time, paid by WeChat or Alipay. It unlocks the full Top 10 in every category, six-dimension comparison scores with radar charts, dark-horse alerts fed by external signals refreshed every 3 hours, CSV export of the rankings, 30 AI searches and task scans a day and 100 AI images a month, plus the full version of each trend report.
Testing a tool from the region it is built for
One of NavXD's five scoring dimensions is availability from mainland China, which is the site's way of saying that a good share of the tools it tracks are easier to reach from elsewhere. A radar tells you a tool exists; the next step is to open it, sign up and run your own task through it.
A PROXIES.SX 4G/5G mobile or residential IP from the country where a tool is offered handles the registration, the verification step and the trial that follows, and a team that evaluates tools for several markets keeps a separate country IP per account. Bandwidth is billed per GB and the GB never expire, which suits an evaluation habit that comes in bursts.
Who NavXD suits
Teams tracking the Chinese AI ecosystem, where the radar covers Kimi, DeepSeek, Doubao, Jimeng, Tongyi, Coze and their peers alongside ChatGPT, Claude and the agent frameworks. Builders who would rather start from a task and a route than from a list of names. And readers who want a weekly-calibrated ranking with a stated set of dimensions and the option to export it.