jichangtz-tuijian: a Chinese list of proxy subscription providers, not software

The repository holds one README and a folder of screenshots

The repository is not a program. Its top level contains a README file and an `image/` directory, and that is the whole of it. There is no source directory, no build script, no test suite and no package manifest, so there is nothing to build and nothing to install. What you consume is a document. The published site at `https://myikuns.github.io/jichangtuijian` renders the same content, and the repository exists so that the content is versioned, reviewable in pull requests, and easy to update from a phone when a provider changes its prices. That is a sensible reason to host a list on GitHub rather than on a personal blog, and it also means the update mechanism is the pull request. The screenshots under `image/` carry the plan tables, the topology diagrams and the evening peak speed tests for each provider, which is where most of the factual detail lives beyond the text. The single file structure also means the quality of what you read depends entirely on how carefully the author maintains one long document.

Three tiers, sixteen providers, one price column

The list is organised in three tiers, and the tiering is a judgement about price and reliability rather than about features. The top tier holds four entries, the middle tier six, and the budget tier six, so sixteen providers are named in total. The top tier mixes high traffic allowances with premium lines: the first entry is listed at 114 yuan for 500GB with 250 or more high speed lines, the second at 189 yuan yearly for 100GB a month with an enterprise grade dedicated line on its higher tier, the third at 66 yuan a quarter for 100GB with IEPL dedicated capacity, and the fourth at 50 yuan for 800GB on a Japanese KDDI residential line. The middle tier starts at 20 yuan for 100GB and includes one entry advertised as unlimited for 9.9 yuan a month. The budget tier is where the numbers get very small, with entries at 2 yuan for 128GB, 6 yuan for 1000GB, 2 yuan for 100GB, 1.5 yuan for 100GB and 1 yuan for 100GB. The traffic allowances are the only part of the list that a reader can compare directly across entries, and even there the units differ, since some prices are per month, some per quarter and some per year.

The signup links carry referral parameters

Every provider link in the tables is a signup link rather than a plain homepage, and this is the part of the repository a reader should notice before clicking anything. Several links carry an invite code in the path, one carries an auth fragment, and some carry a query parameter that looks like a referral identifier. A discount code is listed separately for one provider at 10 percent off, another entry lists a seasonal code, and a third lists a code tied to the entry's own pricing. That arrangement means the list and the provider both benefit when a reader subscribes through it, which is not a hidden mechanism, but it does shape the list. It explains the ordering, since a provider willing to pay a referral commission is cheaper to recommend, and it explains why some entries are described in more detail than others, with topology diagrams and per carrier peak hour charts included only for the providers the author has the closest relationship with. The list also links to a residential IP proxy service, to a marketplace selling low cost overseas accounts, and to two cryptocurrency exchange referral links, all under a heading about clean IPs and high requirement scenarios.

Protocols, device limits and streaming claims as the comparison axis

Since no code exists, the list compares providers through the attributes a subscriber would otherwise have to check one by one. Protocol labels appear throughout, with Shadowsocks named for one of the top tier entries, and the repository topics and description name the wider vocabulary in use, including anytls, hysteria2, trojan, vless, vmess and shadowsocks, alongside client applications such as Clash, Clash Verge, V2rayN, sing-box, nekobox, Shadowrocket, karing and Loon. Device limits are stated per entry and they vary a lot. The first top tier entry allows at most 10 simultaneous devices, the third allows 5, and several lower tier entries advertise no limit at all, with one noting that abuse will trigger a ban. Node counts and country coverage are given in the same form, from 50 or more regions on one entry to 105 or more nodes across 76 or more countries on another. Streaming and AI service accessing is a repeated claim, naming Netflix, Hulu, HBO, Disney+, ChatGPT, Claude, Gemini and Cursor, but the list presents those claims as provider assertions rather than as its own test results.

The author's own warnings sit at the top of the page

The opening lines are the most honest part of the document, and they are not a recommendation. The author states that updates are infrequent but that each entry is the result of long term personal use, that the current period is unusual and node counts will fluctuate, and that a suggestion is given to keep several providers ready as backup. The dedicated lines and transit routes are described as unstable in duration and as disconnecting frequently. A separate note on one entry warns that its price changes without notice and advises buying early, which is advice that only makes sense if you accept that a published price is not a price. The last push on the repository was on 2026-10-19, and there are no GitHub releases, so there is no versioned history of the list to consult. What you can inspect instead is the commit log of one long file, which is a record of what changed and when, and nothing more.

Reading the list without following the links

The alternative to a curated list is to use the client software directly and treat providers as interchangeable, since the clients named in the repository description, including Clash, V2rayN, sing-box and Shadowrocket, are built around subscription links rather than around one vendor. The difference in approach is what you lose and what you gain. A curated list gives you a starting point, a price list and a note about which entry is currently worth trying, and it saves the search. What it gives you instead of evidence is a claim, and the claim is repeated from the provider rather than measured independently, which is why the evening peak charts exist for a handful of entries and not for all sixteen. There is also a structural difference: a list is a snapshot that a competitor can outlast, while a subscription is a live relationship you can audit by measuring your own throughput and looking at your own node list. The clients are where that measurement happens, and the list is where you decide what to point them at.

What this repository cannot tell you

Several questions a serious buyer would ask are simply not answered anywhere in the document. There is no uptime figure, no measurement method, no statement of which client and which carrier a test was run on beyond the per carrier charts, and no explanation of how a provider is removed or added. There is no licence statement and no indication of what a reader is permitted to do with the content, which is a thin legal position for a document that is freely copied and mirrored. The list also describes itself as containing evaluations, but the evidence behind each evaluation is a set of screenshots and a personal opinion, which is a defensible way to write a recommendation and a weak way to make a claim. A reader who is choosing a provider for a work connection should treat the traffic figures, the country counts and the accessing lists as the vendor's marketing copy, and should measure before paying.

Editorial conclusion

Treat this repository as a price index and a shortlist of names to research, not as a review to trust and not as a tool to install. It is useful if you want the current plan tiers, the traffic allowances and the protocol labels in one place, and it is the wrong thing to depend on if you need verified performance data, a stable recommendation, or a vendor-neutral comparison. Before you follow any of its links, check three things: that the referral parameters in the signup URLs do not change the price you pay, that the traffic allowance and protocol claims match what your own client supports, and that the provider you pick is one you can leave when the node list changes. The list was last pushed on 2026-10-19, it has no tagged releases, and it states no licence, so treat the pricing in it as a snapshot rather than a contract.

Frequently asked questions

Official sources

1. Issues
2. myikuns/jichangtz-tuijian on GitHub
3. Project website
4. README