Connect in five minutes

Before you start

You receive three different credentials—use each in the correct place:

API Key

sk- value used to authenticate API requests.

API endpoint

https://api.buytoken.work. Send every API request here.

Username and password

Use these to sign in, inspect usage, and change your password. They are not your API key.

Which models you can call depends on your key’s group: the cc group calls Claude models, codex calls GPT models, kimi calls Kimi, and the DeepSeek group calls DeepSeek. Supported providers are listed on the Supported models. Prices and group notes are on the Model pricing page.

Connect Claude models

Use Claude Code with CC Switch to manage providers. Install CC Switch once; Codex below uses the same app.

1. Install Node.js and Claude Code

   Install Node.js 18+ from nodejs.org, then run this in a terminal (PowerShell on Windows, Terminal on Mac/Linux):

   Terminal Copy

   npm install -g @anthropic-ai/claude-code

2. Install CC Switch

   Download the installer for your OS from github.com/farion1231/cc-switch and complete the setup.

3. Add a provider in CC Switch

   Open CC Switch, add a provider, choose “Custom configuration”, and fill in:

   Name

   Any name, such as “My relay”

   Endpoint

   https://api.buytoken.work (if it does not work, append /v1 and retry)

   API Key

   Your sk- key

   API format

   Anthropic

   Use this model mapping, then save and enable the provider:

   Default

   claude-sonnet-5

   Opus

   claude-opus-5

   Haiku

   claude-haiku-4-5-20251001

4. Start using it

   Run claude in the terminal to start.

   VSCode, Cursor, and Trae users can install the Claude Code for VSCode extension—no sign-in or extra setup required.

Connect GPT · Codex

To use gpt-5.6-sol models, follow this section after installing CC Switch above.

1. Install Codex CLI or Codex Desktop

   Both clients share one configuration. Install either or both; the CLI requires Node.js 18+:

   Terminal Copy

   npm install -g @openai/codex

   For a graphical interface, download the official desktop app from openai.com/codex.

2. Add a Codex provider in CC Switch

   Open CC Switch, select the Codex tab, add a provider, and choose “Custom configuration”:

   Name

   Any name, such as “buytoken-codex”

   Endpoint

   https://api.buytoken.work/v1 (/v1 is required)

   API Key

   Your sk- key

   API format

   OpenAI

   After saving and enabling it, CC Switch writes the endpoint and key to ~/.codex/config.toml; both CLI and desktop use this file.

3. Start using it

   Run codex to enter the interactive UI and use /model gpt-5.6-sol to switch models. In the desktop app, enter the model ID when creating a conversation.

   If the client still uses old settings after a CC Switch change, restart it.

Generate images with gpt-image

Codex’s built-in image_gen is a capability hosted by OpenAI and does not use our endpoint, and its fallback script only accepts b64_json while our image API stores the result and returns a link, so neither path works here. To generate images from Codex, switch the model to gpt-image-2: the relay treats the request as image generation, saves the image under ~/.codex/generated_images/, and also puts the image link in the reply text.

Add a profile to ~/.codex/config.toml; your existing chat model is unaffected:

[model_providers.buytoken]
name = "buytoken"
base_url = "https://api.buytoken.work/v1"
wire_api = "responses"
requires_openai_auth = false
# Or delete the line below and use env_key = "BUYTOKEN_API_KEY" instead
experimental_bearer_token = "sk-your-key"

[profiles.image]
model = "gpt-image-2"
model_provider = "buytoken"

Start the session with --profile; single-shot is the most reliable (Codex uses only the last message as the prompt):

codex --profile image
codex exec --profile image "a ginger cat, realistic photography"

If a session is already open, you can also run /model gpt-image-2.

Codex does not send a size or quality, so output is fixed at 1024x1024 and the upstream default quality. To control those, call the image API directly; data[0].url in the response is the image:

curl -sS https://api.buytoken.work/v1/images/generations \
  -H "Authorization: Bearer sk-your-key" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-image-2","prompt":"a red bicycle","size":"1536x1024","quality":"low","n":1}'

Available models: gpt-image-2, gpt-image-2.5-flare, gpt-image-2.5-sunburst, all billed per image. Generation takes 30–120 seconds, so do not set a short client timeout; a disconnect part-way is still billed. For batches, call the API directly.

Connect WorkBuddy

WorkBuddy is Tencent's AI workspace for office tasks. You can add a third-party model in its graphical settings, with no config file to edit.

Before you start, make sure WorkBuddy is installed and signed in, and that you have a key starting with sk-, and that the key's group includes the model you want. The model name you enter must match the model ID in that group exactly. Look it up in Supported models or Models and pricing.

1. Open model settings

   Click the account area at the bottom left to open the menu, choose Settings, then choose Models in the left navigation.

   

   

2. Add a custom model

   In the Custom models area, click Add model at the top right, open the Provider list, and pick Custom under Other.

   

   

3. Fill in the model settings

   API endpoint

   https://api.buytoken.work/v1 (/v1 is required)

   API Key

   Your sk- key

   Model name

   A model ID from the key's group, for example cc group: claude-opus-4-8

   Advanced options

   Tick what the model supports. For Claude models you can enable tool calling, thinking mode, and allow turning thinking off; leave the rest at their defaults.

   Click Save at the bottom right. WorkBuddy writes the settings to ~/.workbuddy/models.json.

   

4. Start using it

   Go back to a task conversation, open the model selector at the bottom right of the input box, choose the model you just added, and send a message. A reply means it works.

   

   Do not share your key or show it in public. Cover it completely before posting a screenshot or recording to ask for help.

Alternative: environment variables

If you prefer the command line, two environment variables are enough.

Mac / Linux

Edit ~/.zshrc (bash users: ~/.bashrc), append these two lines, save, then run source, then run claude:

export ANTHROPIC_BASE_URL=https://api.buytoken.work
export ANTHROPIC_AUTH_TOKEN=sk-your-key

Windows

Run these two lines in PowerShell (persistent). Close and reopen PowerShell afterward or the changes will not apply.

[System.Environment]::SetEnvironmentVariable("ANTHROPIC_BASE_URL", "https://api.buytoken.work", "User")
[System.Environment]::SetEnvironmentVariable("ANTHROPIC_AUTH_TOKEN", "sk-your-key", "User")

OpenAI-compatible clients

OpenAI-compatible tools such as Codex require /v1:

export OPENAI_BASE_URL=https://api.buytoken.work/v1
export OPENAI_API_KEY=sk-your-key

Verify the setup

Enter claude and send any message. A response confirms the setup. You can also use curl; returned JSON means success:

curl -sS https://api.buytoken.work/v1/messages \
  -H "x-api-key: sk-your-key" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{"model":"claude-sonnet-5","max_tokens":16,"messages":[{"role":"user","content":"hi"}]}'

The gateway supports /v1/messages, /v1/responses and /v1/chat/completions; pass the key with x-api-key or Authorization: Bearer.

Using the console

Open https://api.buytoken.work/console in a browser and sign in with your account credentials to access:

Overview

See balance, monthly spend, and recent requests at a glance.

My Keys

View, copy, rename, disable, and create keys.

Usage logs

Inspect time, model, token counts, and charge for every request; open a row to see the calculation.

Account settings

Change your password; we recommend doing this after first sign-in.

After first sign-in, change the initial password, confirm your requests appear in Usage logs, and check your balance. Top up from Wallet using WeChat Pay.

Troubleshooting

# Mac / Linux
rm -rf ~/.claude

# Windows PowerShell
Remove-Item "$env:USERPROFILE\.claude" -Recurse -Force

Set-ExecutionPolicy -Scope CurrentUser -ExecutionPolicy RemoteSigned

npm config get prefix

{
  "claudeCode.preferredLocation": "panel",
  "claudeCode.environmentVariables": [
    {"name": "ANTHROPIC_BASE_URL", "value": "https://api.buytoken.work"},
    {"name": "ANTHROPIC_AUTH_TOKEN", "value": "sk-your-key"}
  ]
}

"CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS": "1"