The OpenNetCN clash repository, a measurement page for proxy subscriptions

A recommendation page, not the proxy client

The first thing to understand about this repository is that it is not the Clash implementation. The repository contains two files at the top level, `README.md` and `daily-news.md`, and it has no source code, no build system and no license field. Its topics list mentions clash, proxy, sing-box, v2ray and vpn, and its default branch is `main`.

What it is, in the project's own framing, is a recommendation and measurement page for users looking for a stable proxy service, Clash clients, node subscriptions and ways to reach AI tools. The heading of the README describes it as a 2026 roundup of Clash providers, node subscriptions and tested connections, and the page is aimed at people who want a quick answer to whether such a service suits them before they check their devices and target platforms.

The commercial nature is explicit. Every provider row in the summary table ends with a registration link carrying an affiliate or referral code, and the page points to a companion site for the longer reviews and image walkthroughs.

The audience is a reader trying to choose, not a developer integrating something. There is nothing to install, no configuration file to write and no library to depend on. The value on offer is the comparison tables, the availability snapshots and the explanations of how clients and subscriptions work.

That framing also means the article you are reading cannot tell you how to adopt this project, because there is nothing to adopt. What can be assessed is whether the page is a reliable guide and what it actually measures.

How the availability snapshot is put together, and what it cannot tell you

The page includes a daily node availability observation table, and the section explaining how to read it is unusually careful, which is the most interesting thing in the repository.

The snapshot is dated 2026-10-15 and covers all eight providers, with a column for the client used for the check, the last check date, whether the provider was usable, the total node count, the available node count, the available node ratio, a historical available node ratio and a stability rating. The definition is given precisely: the available node ratio is the available node count divided by the total node count for that check, rounded to two decimal places, and the total node count is not the same as the number of nodes every plan can actually use.

Then come the caveats, and they are the part worth reading. The table does not include the full region, carrier, test window, client version, probe target, timeout or retry records, so it cannot be used to reproduce the measurement, and it cannot be used to compare everyone's actual experience. The historical ratio and stability rating are carried over from the provided check records without the start and end dates of the statistics, the number of valid samples, the aggregation algorithm or the rating thresholds, so no long-term conclusion should be drawn from them.

The page is explicit about the biggest trap: a connectivity snapshot is not an all-day online rate, not a speed, not an AI login success rate and not a measure of whether a streaming service will actually let you watch.

That last distinction is the one to carry into any decision. A provider can show 100% of nodes reachable in one snapshot and still be unusable for the specific service you need, because a probe that resolves a node is not the same as a probe that completes a login flow on a platform with its own regional rules.

Clash-compatible clients versus official clients, the practical filter

If you take one operational rule from the page, it is the one stated as a check-before-you-pay instruction: confirm the devices first, then pay.

The compatibility split is specific. WgetCloud, BoostNet and YouTu support general-purpose clients such as Clash, and BoostNet and YouTu also offer their own official clients. The historical records for TAG note support for both official and general clients, with the current compatibility list and versions to be checked against the provider's own site. QingYunTi, YinHeYun, WeiTuYun and TNTCloud currently support only their official clients.

That distinction changes what you can do. A provider that only offers an official client gives you a closed application you cannot point at your own configuration, cannot mix with other tools and cannot inspect. A provider with a Clash-compatible subscription gives you a YAML file you can import into Mihomo, which is the maintained successor line of the Clash family, and which the page links for configuration and subscription import instructions alongside a glossary of the terms.

There is a second piece of advice about combinations that is worth repeating because it is the kind of thing a recommendation page usually glosses over. The page says explicitly that choosing two providers does not mean the underlying lines are independent, and that when the primary fails, whether the backup can still connect needs actual verification. It recommends testing with the short-cycle plans each provider offers first, and warns that some plans are annual-only, that a short-cycle test is not a complete validation of an annual plan's line permissions, and that one provider's 200GB is a total for the whole year rather than a monthly allowance.

There is also a note on how to read the speed and region-access images: the dates in the images are the dates of the samples, the regions, carriers, bandwidth, thread counts and test tools differ, and the region-access label is a tool probe tag that does not guarantee the actual account or all features work.

Purchase terms differ more than the prices do

The provider summaries carry a lot of restrictions that a price list alone would hide, and the page is explicit that the price in the table is a record from the author's own checking rather than a current checkout amount.

The plan structures are unusual. The light annual plan at one provider is priced at a yearly figure with an implied monthly cost, and it is annual only. The lowest-cost trial tiers are monthly and small, with a larger main tier after that, and a year-on-year tier where the annual figure works out to a low monthly equivalent. Another provider's lowest tier is a high yearly price for a small monthly allowance, which is only sold annually and whose refund terms need checking.

Then the restrictions. One provider is limited to mainland China only, with Xinjiang and overseas addresses excluded, and no refunds. Another is unavailable in Xinjiang, in Hong Kong, Macau and Taiwan and overseas. The largest-node provider in the table, with 304 nodes in the snapshot, is historically priced at 162 per year for 200GB across the whole year, or 185 per quarter for 250GB a month, with the same regional exclusions and no guarantee that TikTok works.

The usage rules are worth noting as well. One provider's terms forbid sharing the account with others, forbid BT, P2P and PT traffic, and state that refunds are not provided. Traffic resets are on the purchase date according to the recorded materials, and the actual reset date is whatever the provider's own panel says.

The pattern across the table is that the differences that decide whether a provider is usable for you are the client support, the geographic exclusions and the refund terms, and the price is last on the list. The page's own ordering puts the device check ahead of the purchase, and that ordering is the correct one.

What the page documents about Mihomo and subscription import

The one genuinely technical part of the repository is the section on client configuration and subscriptions, and its table of contents shows what it covers: client downloads, Mihomo configuration, subscription import, a glossary, and a section comparing a proxy subscription service with a VPN.

The glossary in particular is the part that helps a reader who has never configured a proxy client, since the vocabulary in this space is dense and the page defines the terms it uses rather than assuming them.

The page also has a beginner section, a pre-purchase checklist, and a section explaining what a proxy subscription service is alongside one comparing it with a VPN. That last comparison is the one worth noting, because the two are frequently conflated and the page treats them as different things with different properties rather than as synonyms.

The repository also carries a `daily-news.md` file, which is where the daily observation records are kept, separate from the main README.

For a reader who wants to go further, the page points to a companion site with the longer per-provider reviews and image walkthroughs, noting that it requires a working external connection to reach, which is an accurate description of the chicken-and-egg problem anyone in this situation runs into.

What the disclaimer and the risk section actually say

A recommendation page that does not state its own risks is worth little, and this one has a risk section and a disclaimer among its table of contents entries, plus a short list of external resources and a contribution section.

The evidence section states the basis for the recommendations: a long-running personal subscription and comparison across more than forty providers since 2022, with eight kept. The selection criteria are the compatibility of common devices, continuous use during the evening peak, access to target platforms, payment terms and failure-handling records. Each piece of evidence is scoped: the plan and restriction information is compiled from existing official site material with the check dates recorded per brand, the usage observations include suitable scenarios, restrictions and screenshots, and the node observation only shows node counts and availability ratios from existing snapshots and is not used to extrapolate an all-day online rate or whether a platform will actually let you through.

There is a historical observation about an April 2026 outage period, noting that three providers had relatively better recovery and support contact. The page is careful about what that does and does not mean, saying it attaches no complete recovery-time records for a horizontal ranking and that the observation does not constitute a promise about future availability or support response.

That careful scoping is the strongest signal about the page as a whole. It repeatedly says what its data cannot support. A recommendation page that tells you exactly where its evidence stops is more trustworthy than one that presents a single availability percentage as a verdict.

The counterweight is that the page is still an advertisement with referral links, and a reader should treat the provider list as a lead rather than a decision. The page says as much itself in places, noting that the top update time marks the revision of the article and not a re-verification of all prices and client policies on that day, and that the plan check dates appear in each brand summary while node observations and speed screenshots carry their own separate dates.

Editorial conclusion

This repository is worth reading if you want to understand how proxy subscription services are compared and what the measurement caveats are, and it is a reasonable reference for the Mihomo configuration and subscription import mechanics it documents. It is not a software project, so there is nothing to install, no code to audit and no API to integrate. Verify first that the snapshot dates in the tables are recent enough for your own decision, since the page states the top update time marks the revision and not a recheck of every price and client policy, and treat the affiliate links as what they are when you weigh the provider list.

Frequently asked questions

Official sources

1. Issues
2. OpenNetCN/clash on GitHub
3. Project website
4. README