不管你最终使用哪个客户端，接入流程都是一样的三步。整个过程通常不超过五分钟。

Whichever client you end up using, the setup is the same three steps — and it usually takes less than five minutes.

byecore 同时兼容 OpenAI 与 Anthropic 两套 API 规范，且两套协议完全互通——网关自动转换请求格式，控制台中的任意模型都可以通过任一协议调用。统一入口为 https://api.byecore.com（不带任何后缀），不同协议的客户端按下表填写 Base URL：

byecore is compatible with both the OpenAI and Anthropic API specifications, and the two protocols are fully interoperable — the gateway converts request formats automatically, so any model in the Console can be called over either protocol. The unified endpoint is https://api.byecore.com (no suffix); clients on different protocols fill in the Base URL per the table below:

统一入口

Unified endpoint

https://api.byecore.com

OpenAI 兼容客户端

OpenAI-compatible clients

https://api.byecore.com/v1（多数客户端要求带 /v1；Cherry Studio 等会自动补全的只填根地址）

https://api.byecore.com/v1 (most clients require the /v1 suffix; clients that auto-complete it, such as Cherry Studio, take only the root address)

Claude 协议客户端

Claude-protocol clients

https://api.byecore.com（Claude Code 等会自动追加 /v1/messages）

https://api.byecore.com (Claude Code and similar append /v1/messages automatically)

鉴权统一通过 HTTP Header 传递：

Authentication is always passed via an HTTP header:

# OpenAI 协议客户端
Authorization: Bearer sk-你的密钥

# Anthropic 协议客户端
x-api-key: sk-你的密钥

# OpenAI-protocol clients
Authorization: Bearer sk-your-key

# Anthropic-protocol clients
x-api-key: sk-your-key

byecore 聚合多家主流模型厂商，统一入口、统一鉴权、统一计费。具体在售模型与版本以 控制台 实时列表为准，也可以随时调用 /v1/models 查询当前 Key 可用的全部模型。

byecore aggregates many leading model providers behind one endpoint, one set of credentials, and one billing system. The exact models and versions on offer follow the live list in the Console; you can also call /v1/models at any time to see every model available to the current Key.

当你需要在多个路由、多个 Key、多个模型之间频繁切换时，手动编辑配置文件容易出错。开源社区已经有一批成熟的切换与代理工具，按需选择其一即可。

When you frequently switch between multiple routes, keys, and models, editing config files by hand is error-prone. The open-source community already offers a set of mature switching and proxy tools — pick whichever fits your needs.

Endpoint (Codex):       https://api.byecore.com/v1
Endpoint (Claude Code): https://api.byecore.com
API Key:                sk-你的密钥
Default Model:          按控制台列表选择（示例 gpt-5.4）

Endpoint (Codex):       https://api.byecore.com/v1
Endpoint (Claude Code): https://api.byecore.com
API Key:                sk-your-key
Default Model:          pick from the Console list (e.g. gpt-5.4)

# 1. 全局安装（需要先安装 Node.js 与 Claude Code）
npm install -g @musistudio/claude-code-router

# 2. 完成下方配置后，用 ccr 启动 Claude Code
ccr code

# 1. Install globally (Node.js and Claude Code required first)
npm install -g @musistudio/claude-code-router

# 2. After the config below, start Claude Code with ccr
ccr code

{
  "Providers": [
    {
      "name": "byecore",
      "api_base_url": "https://api.byecore.com/v1/chat/completions",
      "api_key": "sk-你的密钥",
      "models": ["gpt-5.4", "gpt-5.5", "gpt-5.3-codex"]
    }
  ],
  "Router": {
    "default": "byecore,gpt-5.4",
    "background": "byecore,gpt-5.4",
    "think": "byecore,gpt-5.5",
    "longContext": "byecore,gpt-5.4"
  }
}

model_list:
  - model_name: gpt-5.4
    litellm_params:
      model: openai/gpt-5.4
      api_base: https://api.byecore.com/v1
      api_key: sk-你的密钥
  - model_name: gpt-5.5
    litellm_params:
      model: openai/gpt-5.5
      api_base: https://api.byecore.com/v1
      api_key: sk-你的密钥

pip install 'litellm[proxy]'
litellm --config config.yaml --port 4000

# 客户端 Base URL 填: http://localhost:4000/v1

pip install 'litellm[proxy]'
litellm --config config.yaml --port 4000

# Set the client Base URL to: http://localhost:4000/v1

以下客户端均可通过「OpenAI 兼容 / 自定义 Base URL」或 Anthropic 协议接入 byecore。卡片角标对应详细配置的小节编号，点击即可跳转。

Every client below can connect to byecore via an OpenAI-compatible / custom Base URL or the Anthropic protocol. The badge on each card is the section number for its detailed config — click to jump there.

Codex 接入 byecore 需要两个文件：config.toml 放模型与服务商配置， auth.json 放 API Key。推荐先用 CC Switch 自动写入；以下为手动配置的完整内容。

Connecting Codex to byecore takes two files: config.toml holds the model and provider config, and auth.json holds the API Key. We recommend using CC Switch to write them automatically; the full manual config follows.

model_provider = "OpenAI"
model = "gpt-5.4"
review_model = "gpt-5.4"
model_reasoning_effort = "xhigh"
disable_response_storage = true
network_access = "enabled"
windows_wsl_setup_acknowledged = true
model_context_window = 1000000
model_auto_compact_token_limit = 900000

[model_providers.OpenAI]
name = "OpenAI"
base_url = "https://api.byecore.com/v1"
wire_api = "responses"
requires_openai_auth = true

{
  "OPENAI_API_KEY": "sk-你的密钥"
}

byecore 原生提供 Anthropic 兼容端点，Claude Code 无需任何中间件即可直连：把 ANTHROPIC_BASE_URL 指向根地址（注意不要带 /v1 后缀），令牌填你的 sk- 密钥。网关自动完成协议转换，控制台中的任意模型都可以在 Claude Code 中使用。

byecore provides a native Anthropic-compatible endpoint, so Claude Code connects directly with no middleware: point ANTHROPIC_BASE_URL at the root address (do not add the /v1 suffix) and use your sk- key as the token. The gateway handles protocol conversion, so any model in the Console works inside Claude Code.

export ANTHROPIC_BASE_URL="https://api.byecore.com"
export ANTHROPIC_AUTH_TOKEN="sk-你的密钥"

claude

$env:ANTHROPIC_BASE_URL   = "https://api.byecore.com"
$env:ANTHROPIC_AUTH_TOKEN = "sk-你的密钥"

claude

想长期固化配置，写入 Claude Code 的全局设置文件即可：

To make the config permanent, write it into Claude Code's global settings file:

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.byecore.com",
    "ANTHROPIC_AUTH_TOKEN": "sk-你的密钥"
  }
}

Cursor 支持用自有 Key 覆盖官方 OpenAI 接口，步骤如下：

Cursor lets you override the official OpenAI endpoint with your own Key. The steps:

入口

Where

Cursor Settings → Models → API Keys

OpenAI API Key

启用并填入 sk-你的密钥

Enable it and enter sk-your-key

Override Base URL

勾选 Override OpenAI Base URL，填 https://api.byecore.com/v1

Check "Override OpenAI Base URL" and enter https://api.byecore.com/v1

添加模型

Add models

在模型列表中 Add model，手动输入 gpt-5.4、gpt-5.5 等，并只勾选这些自定义模型

Add model in the model list, manually enter gpt-5.4, gpt-5.5, etc., and enable only these custom models

验证

Verify

点击 Verify，通过后即可在 Chat / Composer 中选用 byecore 模型

Click Verify; once it passes, you can use byecore models in Chat / Composer

这三款 VS Code Agent 插件同源同构，配置界面几乎一致：在插件设置中选择 OpenAI Compatible 作为 API Provider，然后填三件套。

These three VS Code agent extensions share the same origin and structure, and their settings UIs are nearly identical: in the extension settings, choose OpenAI Compatible as the API Provider, then fill in the three fields.

API Provider

OpenAI Compatible

Base URL

https://api.byecore.com/v1

API Key

sk-你的密钥

Model ID

gpt-5.4（编程密集型任务可改 gpt-5.3-codex，重分析任务用 gpt-5.5）

gpt-5.4 (use gpt-5.3-codex for coding-heavy tasks, gpt-5.5 for analysis-heavy ones)

Continue 通过 config.yaml 管理模型（旧版为 config.json，字段同名）。打开侧边栏 → 齿轮图标 → Open Config，加入：

Continue manages models via config.yaml (older versions used config.json with the same field names). Open the sidebar → gear icon → Open Config, and add:

name: byecore
version: 0.0.1
schema: v1

models:
  - name: byecore gpt-5.4
    provider: openai
    model: gpt-5.4
    apiBase: https://api.byecore.com/v1
    apiKey: sk-你的密钥
    roles:
      - chat
      - edit
      - apply

  - name: byecore gpt-5.5
    provider: openai
    model: gpt-5.5
    apiBase: https://api.byecore.com/v1
    apiKey: sk-你的密钥
    roles:
      - chat

Aider 是最轻量的接入方式：设置两个环境变量，启动时带上 openai/ 前缀的模型名即可。

Aider is the lightest way to connect: set two environment variables and launch with a model name prefixed by openai/.

export OPENAI_API_BASE="https://api.byecore.com/v1"
export OPENAI_API_KEY="sk-你的密钥"

aider --model openai/gpt-5.4

$env:OPENAI_API_BASE = "https://api.byecore.com/v1"
$env:OPENAI_API_KEY  = "sk-你的密钥"

aider --model openai/gpt-5.4

OpenCode 通过 opencode.json 声明自定义 Provider，底层使用 @ai-sdk/openai-compatible 驱动：

OpenCode declares a custom Provider in opencode.json, driven under the hood by @ai-sdk/openai-compatible:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "byecore": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "byecore",
      "options": {
        "baseURL": "https://api.byecore.com/v1",
        "apiKey": "sk-你的密钥"
      },
      "models": {
        "gpt-5.4": { "name": "gpt-5.4" },
        "gpt-5.5": { "name": "gpt-5.5" },
        "gpt-5.3-codex": { "name": "gpt-5.3-codex" }
      }
    }
  },
  "model": "byecore/gpt-5.4"
}

Zed 在 settings.json（命令面板 → zed: open settings）中配置 OpenAI 兼容接口：

Zed configures an OpenAI-compatible endpoint in settings.json (command palette → zed: open settings):

{
  "language_models": {
    "openai": {
      "api_url": "https://api.byecore.com/v1",
      "available_models": [
        {
          "name": "gpt-5.4",
          "display_name": "byecore · gpt-5.4",
          "max_tokens": 1000000
        },
        {
          "name": "gpt-5.5",
          "display_name": "byecore · gpt-5.5",
          "max_tokens": 1000000
        }
      ]
    }
  }
}

Crush（charmbracelet 出品）通过 crush.json 声明 OpenAI 兼容 Provider：

Crush (by charmbracelet) declares an OpenAI-compatible Provider in crush.json:

{
  "$schema": "https://charm.land/crush.json",
  "providers": {
    "byecore": {
      "name": "byecore",
      "type": "openai",
      "base_url": "https://api.byecore.com/v1",
      "api_key": "sk-你的密钥",
      "models": [
        {
          "id": "gpt-5.4",
          "name": "gpt-5.4",
          "context_window": 1000000,
          "default_max_tokens": 64000
        },
        {
          "id": "gpt-5.5",
          "name": "gpt-5.5",
          "context_window": 1000000,
          "default_max_tokens": 64000
        }
      ]
    }
  }
}

Factory 的 Droid CLI 支持 BYOK（Bring Your Own Key）自定义模型，把 byecore 加进 ~/.factory/config.json：

Factory's Droid CLI supports BYOK (Bring Your Own Key) custom models. Add byecore to ~/.factory/config.json:

{
  "custom_models": [
    {
      "model_display_name": "byecore gpt-5.4",
      "model": "gpt-5.4",
      "base_url": "https://api.byecore.com/v1",
      "api_key": "sk-你的密钥",
      "provider": "generic-chat-completion-api",
      "max_tokens": 64000
    }
  ]
}

桌面与 Web 端对话客户端适合快速验证 Key、对比模型输出，或承担日常问答。注意各家对 /v1 的拼接规则不同。

Desktop and web chat clients are great for quickly verifying a Key, comparing model outputs, or everyday Q&A. Note that each handles the /v1 suffix differently.

图片模型 image-2 不在编程客户端里直接调用，统一通过独立图片站使用，前端基于 gpt_image_playground。文档站与图片站分离，方便各自独立维护。

The image model image-2 is not called directly from coding clients; it is used through a dedicated image site whose frontend is based on gpt_image_playground. The docs site and image site are kept separate so each can be maintained independently.

PREVIEW · 1.png

PREVIEW · 2.png

配置完成后，建议先在终端验证：Key 是否有效（模型列表）、OpenAI 协议对话是否正常返回；使用 Claude 生态客户端的，再验证一次 Anthropic 协议端点。

Once configured, verify in the terminal first: that the Key is valid (the model list) and that OpenAI-protocol chat returns normally; if you use a Claude-ecosystem client, verify the Anthropic-protocol endpoint as well.

# 1. 验证 Key 与可用模型
curl https://api.byecore.com/v1/models \
  -H "Authorization: Bearer sk-你的密钥"

# 2. 发送一条最小对话请求
curl https://api.byecore.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-你的密钥" \
  -d '{
    "model": "gpt-5.4",
    "messages": [{ "role": "user", "content": "你好，介绍一下你自己" }]
  }'

# 3.（可选）验证 Anthropic 协议端点 —— 任意模型同样可用
curl https://api.byecore.com/v1/messages \
  -H "Content-Type: application/json" \
  -H "x-api-key: sk-你的密钥" \
  -H "anthropic-version: 2023-06-01" \
  -d '{
    "model": "gpt-5.4",
    "max_tokens": 64,
    "messages": [{ "role": "user", "content": "你好" }]
  }'

# 1. Verify the Key and available models
curl https://api.byecore.com/v1/models \
  -H "Authorization: Bearer sk-your-key"

# 2. Send a minimal chat request
curl https://api.byecore.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-your-key" \
  -d '{
    "model": "gpt-5.4",
    "messages": [{ "role": "user", "content": "Hello, introduce yourself" }]
  }'

# 3. (Optional) Verify the Anthropic-protocol endpoint — any model works too
curl https://api.byecore.com/v1/messages \
  -H "Content-Type: application/json" \
  -H "x-api-key: sk-your-key" \
  -H "anthropic-version: 2023-06-01" \
  -d '{
    "model": "gpt-5.4",
    "max_tokens": 64,
    "messages": [{ "role": "user", "content": "Hello" }]
  }'

# 1. 验证 Key 与可用模型
Invoke-RestMethod -Uri "https://api.byecore.com/v1/models" `
  -Headers @{ Authorization = "Bearer sk-你的密钥" }

# 2. 发送一条最小对话请求
$body = @{
  model    = "gpt-5.4"
  messages = @(@{ role = "user"; content = "你好" })
} | ConvertTo-Json -Depth 5

Invoke-RestMethod -Uri "https://api.byecore.com/v1/chat/completions" `
  -Method Post -ContentType "application/json" `
  -Headers @{ Authorization = "Bearer sk-你的密钥" } -Body $body

# 3.（可选）验证 Anthropic 协议端点 —— 任意模型同样可用
$claude = @{
  model      = "gpt-5.4"
  max_tokens = 64
  messages   = @(@{ role = "user"; content = "你好" })
} | ConvertTo-Json -Depth 5

Invoke-RestMethod -Uri "https://api.byecore.com/v1/messages" `
  -Method Post -ContentType "application/json" `
  -Headers @{ "x-api-key" = "sk-你的密钥"; "anthropic-version" = "2023-06-01" } `
  -Body $claude

# 1. Verify the Key and available models
Invoke-RestMethod -Uri "https://api.byecore.com/v1/models" `
  -Headers @{ Authorization = "Bearer sk-your-key" }

# 2. Send a minimal chat request
$body = @{
  model    = "gpt-5.4"
  messages = @(@{ role = "user"; content = "Hello" })
} | ConvertTo-Json -Depth 5

Invoke-RestMethod -Uri "https://api.byecore.com/v1/chat/completions" `
  -Method Post -ContentType "application/json" `
  -Headers @{ Authorization = "Bearer sk-your-key" } -Body $body

# 3. (Optional) Verify the Anthropic-protocol endpoint — any model works too
$claude = @{
  model      = "gpt-5.4"
  max_tokens = 64
  messages   = @(@{ role = "user"; content = "Hello" })
} | ConvertTo-Json -Depth 5

Invoke-RestMethod -Uri "https://api.byecore.com/v1/messages" `
  -Method Post -ContentType "application/json" `
  -Headers @{ "x-api-key" = "sk-your-key"; "anthropic-version" = "2023-06-01" } `
  -Body $claude