English
한국어 English 日本語 中文

Install once,
connect anywhere

Register accounts in the CLI, create a route URL, then put Base URL and sk-gt in whatever tool you use. Check remaining quota with llms status --refresh.

curl -fsSL https://llms.goodtek.xyz/install.sh | bash

1. Install the CLI

Linux / macOS, amd64 or arm64. This page and the script are public. CLI binaries live on private GitHub Releases, so you need gh login or GH_TOKEN.

Recommended

# gh auth status OK
curl -fsSL https://llms.goodtek.xyz/install.sh | bash

Token only

export GH_TOKEN=ghp_xxxxxxxx
curl -fsSL https://llms.goodtek.xyz/install.sh | bash

Binary only

OS=$(uname -s | tr '[:upper:]' '[:lower:]')
ARCH=$(uname -m | sed 's/x86_64/amd64/;s/aarch64/arm64/')
gh release download v0.1.0 -R goodtekxyz/llms -p "llms_${OS}_${ARCH}"
chmod +x "llms_${OS}_${ARCH}"
sudo mv "llms_${OS}_${ARCH}" /usr/local/bin/llms

Verify: llms --help

2. Log in

llms login

Approve the GitHub device code; sk-gt is stored in ~/.config/llms/credentials.json. Default API base is https://llms.goodtek.xyz.

llms status

3. Accounts & routes

llms add
llms route create
llms route key <slug>
llms status --refresh

Routing presets

failoverSequential. If #1 fails, try #2
balanceEven spread (round_robin)
prefer-primaryPrimary weighted 80/20
quota-firstPrefer higher remaining % (quota_aware)
parallelAPI-key race · OAuth excluded by default

Model list: llms models <slug> (real model ids, not account refs)

4. Check remaining quota

Cached only: llms status. Re-read Codex/Claude OAuth usage: llms status --refresh. 
llms status --refresh

The board shows account health (●/○/✖), quota …% bars, reset times, and route strategies. 429s appear as cooldown and leave the pool briefly.

5. Attach a client

You do not need to know which app you will use. Any tool that accepts an OpenAI-compatible (or Anthropic Messages) Base URL works the same.

SettingValue
Base URLhttps://llms.goodtek.xyz/r/<slug>/v1
API Keyllms route key → sk-gt-…
 
curl -sS https://llms.goodtek.xyz/r/<slug>/v1/chat/completions \
  -H "Authorization: Bearer sk-gt-…" \
  -H "Content-Type: application/json" \
  -d '{"model":"<model-id>","messages":[{"role":"user","content":"ping"}],"max_tokens":16}'

For shells: llms env <slug>

Machine-readable source

Automations and agents should read markdown, not HTML.

curl -fsSL https://llms.goodtek.xyz/install.md
curl -fsSL https://llms.goodtek.xyz/control/v1/meta