Refresh every entry against live sources as of Sep 29, 2026. Remove
GitHub Models, xAI, LongCat, Volcengine Ark free tier, Baidu Qianfan and
Empero, which no longer offer a free tier or are offline. Add Volcengine
Ark Coding Plan, Atlas Cloud, NanoGPT, Tencent TokenHub, Alibaba free
quota, Z.ai free Flash models, Nebius Token Factory and Novita. Replace
footnote check-dates with a closing Checked line and document that in
CLAUDE.md.
Adds LongCat, SenseNova, AMD Token Factory, Volcengine Ark, Baidu Qianfan,
iFlytek Spark, AIHubMix, OVHcloud AI Endpoints, LLM7.io, Token Harbor,
Aion Labs, Experiential Labs, and Empero to the free providers section,
each with limits, endpoint, source, and check date.
Removes .claude/settings.json; permissions now live in the ignored
settings.local.json.
The Pages project is connected to the repository directly, so an Actions
workflow running wrangler would deploy a second time on every push and
race it. Cloudflare builds itself; only the build command has to carry
the --config flag, which the dashboard supplies.
- remove .github/workflows/cloudflare-pages.yml and its secret setup
- document the required dashboard build command and build system v3
- note that manual wrangler deploys are rejected by Git-connected
projects
Build the site from one source and deploy it to both hosts from GitHub
Actions, so neither depends on a host's default Jekyll toolchain.
- Gemfile/Gemfile.lock pin Jekyll 4.4.1 on Ruby 3.4; both workflows use
them, which keeps the two sites on an identical toolchain
- _config.yml carries the GitHub Pages url and baseurl, fixing asset
paths that previously resolved off-site; _config.cloudflare.yml layers
over it for the pages.dev root
- wrangler.jsonc declares the Pages project name and output directory
- 404.html gives both hosts a page for unmatched routes
- exclude README.md from the output and stop shipping CSS sourcemaps
Point the referral at https://agentkit.best/?ref=BWA910UK, describe the
current Engineer/Marketing kits and bundle pricing, and refresh the checked
date.
- BytePlus standard pricing ($15/$35), Xiaomi MiMo Token Plan, opencode Go model list
- Z.ai/Synthetic/Alibaba/NVIDIA NIM/Google AI Studio/Cerebras/BigModel/Together AI updates
- remove Modal Labs, LongCat AI, Pollinations AI
Chinese (mainland) counterpart of Z.ai with CNY pricing (Lite ¥49,
Pro ¥149, Max ¥469). Documents the bigmodel.cn referral program —
5% friend discount and tiered cashback at 3/30 invitees.
Add CLAUDE.md with project guidance for adding new LLM providers.
Add .claude/memory/ with feedback rules:
- Preserve markdown quotes when editing
- Save memories in project directory
tiennm99