1. Plugin Configuration
beeswitch docs
  • Plugin Configuration
    • API Relay Setup
    • Claude Code for VS Code Tutorial
    • Claude Code Installation and Usage Guide
    • Gemini CLI Configuration and Usage Guide
    • Codex CLI Configuration Guide
    • OpenCode Integration Configuration Guide
    • N8N Workflow Relay API Usage Guide
  • List Models
    • Models (list models)
      GET
  • OpenAI Format (Supports Major Official Provider Models)
    • Chat (Responses)
      • Create Web Search
      • Create Model Response (GPT-5 Reasoning Enabled)
      • Create Function Call
      • Create Model Response
      • Create Function Call
      • Create Model Response (Streaming)
      • Create Model Response (Control Reasoning Length)
    • ChatGPT APIs
      • ChatGPT Audio
        • Create Video
        • Speech-to-Text gpt-4o-transcribe
        • Create translation (not supported)
        • Speech-to-Text whisper-1
        • Speech-to-Text gpt-4o-transcribe
        • Create speech gpt-4o-mini-tts
        • Create translation (not supported)
      • ChatGPT Chat
        • List Models
        • Create Video with Image
        • Continuously modify and generate videos
        • Create chat picture recognition (streaming) best64
        • Official N test
        • Create Structured Output
        • Control Reasoning Effort
        • Create Chat Function Call
        • deepseek-ocr recognition
        • Create Chat Completion (Non-streaming)
      • ChatGPT Completions
        • Create Completion
    • Images
      • Modify pictures (images)
      • Create Chat Completion (Streaming)
      • Create chat completion qwen-mt-turbo
      • Create chat completion deepseek v3.1 thinking level (streaming)
    • Audio
      • Speech recognition(audio)
      • Speech synthesis (audio)
      • Official Function Calling
      • Create a chat creation graph (non-streaming)
    • Embeddings
      • Text Embeddings
  • Anthropic Format
    • Chat
      POST
    • Chat (prompt cache)
      POST
    • Streaming Response
      POST
    • Chat (Deep Reasoning)
      POST
    • Tool Calling (Function Call)
      POST
    • Tool Calling (MCP) - Pending Support
      POST
    • Analyze Image
      POST
  • Google Gemini APIs
    • Native Format
      • Text-to-Image (Aspect Ratio + Clarity Controls)
      • Generate Image
      • Text Generation
      • Text Generation (Streaming)
      • Text Generation + Reasoning (Streaming)
      • Image Generation
      • Formatted Output
      • Function Calling
      • Document Understanding
      • URL context [native format]
      • Code Execution
      • google search
      • Video Understanding
      • URL context
      • Video understanding-url [native format]
      • Imagen 4 under development
      • Audio Understanding
      • Embeddings
      • Chat
      • Edit Image
    • Image-to-Image Base64 Request (Officially Recommended)
      • Multi-Image Fusion Generation with gemini-3-pro-image-preview (Aspect Ratio + Clarity Controls)
      • Image Editing
      • Single Image with gemini-3-pro-image-preview (Aspect Ratio + Clarity Controls)
      • Image Generation with gemini-2.5-flash-image
      • Image Generation with gemini-2.5-flash-image (Aspect Ratio Control)
      • Image Understanding
    • Image-to-Image URL Request / OpenAI-style URL Request
      • Single Image-to-Image with gemini-3-pro-image-preview (Aspect Ratio + Clarity Controls)
      • Multi-Image Fusion Generation with gemini-3-pro-image-preview (Aspect Ratio + Clarity Controls)
      • Image Editing
      • Image Understanding
    • Image-to-Image URL Request (Not Currently Supported)
      • Multi-Image Fusion with gemini-3-pro-image-preview (Aspect Ratio + Clarity Controls)
      • Image Editing
      • Image Understanding
    • Image-to-Image URL Request, Gemini Format (Not Currently Supported)
      • Multi-Image Fusion Generation with gemini-3-pro-image-preview (Aspect Ratio + Clarity Controls)
      • Image Generation with gemini-3-pro-image-preview (Aspect Ratio + Clarity Controls)
      • Image Editing
      • Image Understanding
  • NanoBanana
    • Image-to-Image Base64 Request (Officially Recommended)
      • Multi-Image Fusion Generation with gemini-3-pro-image-preview (Aspect Ratio + Clarity Controls)
      • Image Editing
      • Single Image with gemini-3-pro-image-preview (Aspect Ratio + Clarity Controls)
      • Image Generation with gemini-2.5-flash-image
      • Image Generation with gemini-2.5-flash-image (Aspect Ratio Control)
      • Image Understanding
    • Gemini Request Format
      • Generate Image
      • Edit Image
    • OpenAI Request Format
      • Edit Image
      • OpenAI Image Format
  • Midjourney Format
    • Task Query API
    • Upload Image
    • Get Seed
    • Submit Imagine Task
    • Query Tasks by ID List
    • Face Swap
    • Execute Action
    • Submit Blend Task
    • Submit Describe Task
    • Submit Modal
    • Refresh URL
    • Edit Image
    • Query Task Status by Task ID
    • Get Seed for Task Image
  • Doubao Video Generation
    • Text-to-Video Example
    • Image-to-Video - First Frame
    • seedance-lite - First and Last Frames
    • Image-to-Video - Base64 Encoding
    • seedance-lite - Reference Image
    • Query Single Task
    • Query video generation task list-default
    • Query video generation task list - search multiple task IDs
  • Doubao Image Generation
    • doubao-seededit-3-0-i2i-250628
    • doubao-seedream-4-0-250828 - Text-to-Image
    • doubao-seedream-4-0-250828 - Image-to-Image
    • doubao-seedream-4-0-250828 - Multi-Image-to-Image
  • Jimeng AI
    • Jimeng AI - Image Generation
      • seedream-4.5
      • seedream-4.5 - Text-to-Image
      • seedream-4.5 - Image-to-Image
      • seedream-4.5 - Multi-Image-to-Image
    • Jimeng AI - Video Generation
      • Unified Task Submission API
      • Unified Query API
  • Rerank Models
    • Rerank
  • Text-to-Music (Suno)
    • Task Submission
      • Generate Song (Inspiration Mode)
      • Generate Song (Custom Mode)
      • Generate Song (Continuation Mode)
      • Generate Song (Singer Style)
      • Generate Song (Remix Uploaded Song)
      • Generate Song (Song Concatenation)
      • Generate Lyrics
      • Song Concatenation
    • Query APIs
      • Fetch Tasks in Batch
      • Query Single Task
  • Veo Video Generation
    • Latest Veo Video Generation (Recommended First)
      • Upload Image (Valid for One Day)
      • Create Video Generation Task - Text Only
      • Create Video Generation Task with Reference Image
      • Query Task Status
    • OpenAI Video Format (Recommended)
      • OpenAI Create Video with Image
      • OpenAI Query Task
      • OpenAI Download Video
    • Unified Video Format (Pending Unification)
      • Create Video
      • Create Video with Image
      • Query Task
      • Create Video (Reference Image)
  • Sora Video Generation
    • Official OpenAI Video Format (Recommended)
      • sora-2/sora-2-pro
        • OpenAI Query Task
        • OpenAI Download Video
        • OpenAI Create Video with Image
        • OpenAI Create Video (with Character)
        • OpenAI Edit Video
    • Chat Format
  • Kling (Kuaishou)
    • Text-to-Video
    • Image-to-Video
    • Query tasks (free)
  • Wan (Qwen)
    • Create Video with Image (Wan)
    • Query Video (Wan)
  • MinMax Video Generation
    • Text-to-Video Generation Task (MinMax)
    • Image-to-Video Task (MinMax)
    • Query Video Generation Task Status (MinMax)
    • Video Download (MinMax)
  • Vidu Video Generation
    • Vidu Generate Video
    • Vidu Query
  1. Plugin Configuration

API Relay Setup

Connect BeeSwitch to API Relay#

BeeSwitch gives you one API key for compatible AI models across your coding workflow. This guide shows you how to connect that key to these popular 3rd party opensource API Relay desktop tools to enable dynamic control over API switching/MCP, Skills, Sessions, cross app session management, and more.
CC Switch — the simplest choice if you use both Claude Code and Codex or regularly switch between providers.
Codex++ — a launcher and provider manager for the official Codex desktop app.
Cockpit Tools — a broader local account and multi-instance manager, including Codex API-key accounts.
If you only want our recommendation, start with CC Switch. BeeSwitch includes a one-click CC Switch connection directly on every API-key row, so you do not need to copy the Base URL or manually build the provider configuration.
These are independent, third-party open-source projects. Their screens and menu names may change between releases. Download them only from the official websites or GitHub repositories linked in this guide.

Before you begin#

You will need:
1.
An active BeeSwitch account with available balance or an eligible plan.
2.
A BeeSwitch API key created from the API Keys page in your BeeSwitch dashboard.
3.
At least one model enabled for that key.
4.
The exact model ID you want to use. Copy it from the BeeSwitch model list instead of guessing it.
5.
The underlying app you plan to run, such as Claude Code, Codex CLI, or the official Codex desktop app.

BeeSwitch connection details#

Use caseBeeSwitch Base URLRecommended protocol
Codex and OpenAI-compatible toolshttps://api.beeswitch.com/v1Responses API
Claude Code and Anthropic-compatible toolshttps://api.beeswitch.comAnthropic Messages
For Claude Code, do not add /v1 to ANTHROPIC_BASE_URL. Claude Code adds the /v1/messages path itself. For Codex, use the Base URL ending in /v1.

Option A: CC Switch#

CC Switch manages provider configurations for Claude Code, Codex, and several other coding tools from one desktop interface. It is the best general-purpose choice if you use more than one coding agent or expect to switch providers frequently.
Official links:
CC Switch website
CC Switch GitHub repository
Latest releases

1. Install CC Switch#

Windows: Download the current .msi installer or portable .zip from the official Releases page.
macOS: Download the current .dmg, or install with Homebrew:
Linux: Download the appropriate .deb, .rpm, or .AppImage from the official Releases page.
Install Claude Code or Codex separately before continuing. CC Switch manages their configuration; it is not a replacement for the coding agent itself.

2. Connect BeeSwitch with one click#

After you create your BeeSwitch API key:
1.
Open API Keys in BeeSwitch.
2.
Find the API key you want to use.
3.
Under Actions, select the three-dot ⋯ menu.
4.
Select CC Switch.
1788876403895_d.png
BeeSwitch API Keys page: open the action menu for the selected key, then choose CC Switch.
5.
Allow your browser to open the ccswitch:// link. Your operating system may display a confirmation prompt the first time.
6.
CC Switch will open the provider configuration dialog with your BeeSwitch connection information already filled in.
7.
Choose the target application: Claude, Codex, or another supported application shown in your version.
8.
Enter a recognizable name such as BeeSwitch.
9.
Select the main model and any optional model tiers you want to use.
10.
Select Open CC Switch or confirm the import.
11.
Enable the new BeeSwitch provider.
12.
Restart the target coding agent if it was already running.
Your API key remains hidden in the BeeSwitch table. You do not need to expose or manually copy it when using the one-click connection.

3. Confirm the imported provider#

After the import, CC Switch should show BeeSwitch in its provider list. Select the provider and choose Enable if it is not already active.
CC Switch provider management interface
CC Switch interface shown for orientation. The current layout may vary by release.
Send a small test request in Claude Code or Codex. Then open Usage Logs in BeeSwitch and confirm that the request appears under the expected API key and model.

4. Manual Claude Code setup — fallback only#

Use this section only if the one-click link does not open or the import is interrupted.
1.
Open CC Switch and select Claude Code.
2.
Select Add Provider, then choose a custom provider configuration.
3.
Enter the following values:
FieldValue
NameBeeSwitch
Base URLhttps://api.beeswitch.com
API key / auth tokenYour BeeSwitch API key
Protocol, if shownAnthropic Messages
Main modelAn exact Claude-compatible model ID from BeeSwitch
4.
If CC Switch offers separate Haiku, Sonnet, or Opus fields, select the BeeSwitch model you want for each tier. You may leave optional tiers unassigned.
5.
Save the provider and select Enable.
6.
Close any already-running Claude Code process, open a new terminal, and run:
CC Switch may support hot switching for Claude Code, but reopening the agent is the safest way to avoid a stale session.

5. Manual Codex setup — fallback only#

Use this section only if the one-click link does not open or the import is interrupted.
1.
Open CC Switch and select Codex.
2.
Select Add Provider, then choose a custom provider configuration.
3.
Enter the following values:
FieldValue
NameBeeSwitch
Base URLhttps://api.beeswitch.com/v1
API keyYour BeeSwitch API key
Protocol / Wire APIResponses
ModelAn exact Codex-compatible model ID from BeeSwitch
4.
Save the provider and select Enable.
5.
Restart Codex or open a new terminal before starting codex.
The custom-provider screen normally looks similar to this:
CC Switch Add Provider screen
CC Switch Add Provider screen shown for orientation. Use the BeeSwitch values above rather than any example values visible in the screenshot.
If you switch back to an official provider later, use CC Switch to enable that saved provider instead of deleting your BeeSwitch configuration.

Option B: Codex++#

Codex++ is designed for the official Codex desktop app. It adds provider switching, model management, diagnostics, session tools, and optional interface enhancements without modifying the official app's installed files.
Codex++ Manager interface
Codex++ Manager shown for orientation. The current layout may vary by release.
Official links:
Codex++ GitHub repository
Latest releases

1. Install the required apps#

1.
Install the official Codex desktop app first.
2.
Download the correct Codex++ installer from GitHub Releases:
Windows: CodexPlusPlus-*-windows-x64-setup.exe
macOS Intel: CodexPlusPlus-*-macos-x64.dmg
macOS Apple Silicon: CodexPlusPlus-*-macos-arm64.dmg
3.
Open Codex++ Manager.
4.
Confirm that the manager detected the official Codex app. If it did not, select the app path manually.

2. Create a BeeSwitch provider#

1.
In Codex++ Manager, open the provider configuration area.
2.
Select Add provider.
3.
Enter the following values:
FieldValue
Provider nameBeeSwitch
Access modePure API
Base URLhttps://api.beeswitch.com/v1
API keyYour BeeSwitch API key
ProtocolResponses
Test modelAn exact Codex-compatible model ID from BeeSwitch
4.
Use model discovery, if available, to load the models accessible to your key.
5.
Save the provider.
6.
Switch the active provider to BeeSwitch.
Why Pure API? It keeps your BeeSwitch key and provider configuration separate from an official ChatGPT/Codex login. Use Official login + API only if you specifically understand and need Codex++'s mixed mode.

3. Test and launch#

1.
Run the provider's Model test or Provider Doctor.
2.
Confirm that the Base URL, API key, protocol, and test model all pass.
3.
Close any existing official Codex desktop process.
4.
Launch the app through the Codex++ entry, not directly from the original Codex shortcut.
Codex++ applies its saved BeeSwitch profile during launch. Opening the official app directly can bypass that profile.

4. If a model only supports Chat Completions#

Use Responses whenever the selected BeeSwitch route supports it. If the route supports only Chat Completions, change the provider protocol to Chat Completions. Codex++ can convert Chat Completions locally into the Responses format expected by Codex, but Codex must then be launched through Codex++ for the local protocol bridge to run.

Option C: Cockpit Tools#

Cockpit Tools is a local desktop manager for multiple AI coding accounts and parallel app instances. The steps below use its Codex API-key workflow with BeeSwitch.
Cockpit Tools Codex accounts interface
Cockpit Tools Codex accounts page shown for orientation. The current layout may vary by release.
Official links:
Cockpit Tools GitHub repository
Latest releases

1. Install Cockpit Tools#

Windows: Download the current .msi or .exe from GitHub Releases.
macOS: Download the correct .dmg, or install with Homebrew:
Linux: Download the appropriate .deb, .rpm, or .AppImage from GitHub Releases.
Cockpit Tools' repository currently describes its default license as non-commercial. If you plan to use it inside a business or as part of a paid service, review the current repository license and contact its maintainer if required.

2. Add BeeSwitch as a model provider#

1.
Open Cockpit Tools and go to Codex.
2.
Open the Model Providers tab.
3.
Select Add Provider.
4.
Choose Custom and enter:
FieldValue
Provider nameBeeSwitch
Base URLhttps://api.beeswitch.com/v1
API keyYour BeeSwitch API key
Protocol, if shownResponses
5.
Select Fetch from Provider under Model List.
6.
Review the returned model IDs and remove any models you do not want displayed.
7.
Enable Sync Provider Models to Codex if you want Cockpit Tools to generate and manage Codex's model catalog.
8.
Save the provider.

3. Add and activate the BeeSwitch API-key account#

1.
Return to the Codex accounts page.
2.
Select Add Account.
3.
Choose API Key.
4.
Reuse the saved BeeSwitch provider, or enter the BeeSwitch Base URL and API key when prompted.
5.
Select the default model.
6.
Save and switch to the new account.
7.
Launch Codex from Cockpit Tools.
Cockpit Tools may start third-party providers through a local gateway for instance isolation and protocol conversion. This is expected. If you use its multi-instance feature, bind the intended BeeSwitch API-key account to each instance before launching it.

4. Optional Claude CLI connection#

Some Cockpit Tools releases also expose an Add to Claude CLI action for compatible providers. If it is available, use:
FieldValue
Base URLhttps://api.beeswitch.com
API keyYour BeeSwitch API key
ProtocolAnthropic Messages
ModelAn exact Claude-compatible model ID from BeeSwitch
If your installed version does not show this action, use CC Switch for Claude Code instead.

Verify your BeeSwitch key independently#

If a desktop tool reports a vague connection error, test the key outside the tool. These examples use environment variables so the key is not written into your terminal history as part of the command.

macOS or Linux#

Windows PowerShell#

$env:BEESWITCH_API_KEY = Read-Host "Paste your BeeSwitch API key"

List available models#

Test the Responses API#

Replace <MODEL_ID> with an exact model ID returned by the model-list request:

Test the Anthropic Messages API#

Replace <CLAUDE_MODEL_ID> with an exact Claude-compatible model ID:

Troubleshooting#

ProblemWhat to check
401 UnauthorizedThe key may be incomplete, revoked, disabled, or pasted into the Base URL field. Create a new BeeSwitch key if necessary.
403 ForbiddenCheck the key's quota, expiration, allowed models, group, IP restrictions, and your BeeSwitch balance.
404 Not FoundCheck the Base URL and protocol. A common mistake is producing /v1/v1/... or using the Claude root URL in a Codex provider.
model_not_foundCopy the exact model ID from BeeSwitch. Confirm that the key's group is allowed to use it.
Model list works, but chat failsThe selected route may not support the tool's required protocol or features. For Codex, prefer a route that supports the Responses API and tool calling.
Changes do not take effectFully quit and reopen the coding agent. For Codex++ and Cockpit Tools, launch Codex from the manager that applied the provider profile.
Tool calls or streaming failTry another compatible model route and confirm that the provider is using Responses for Codex or Anthropic Messages for Claude Code.
Request appears in BeeSwitch but the app failsReview the BeeSwitch request log for the upstream error, then run the desktop tool's connection test or diagnostics.
No request appears in BeeSwitchThe app is probably still using an old or official provider. Re-enable BeeSwitch and restart the app.

Key security recommendations#

Create a separate BeeSwitch API key for each tool or device. This makes usage easier to audit and lets you revoke one integration without interrupting the others.
Give each key a reasonable quota and expiration date.
Never post a real key in GitHub issues, screenshots, logs, chat messages, or shared configuration files.
Do not commit auth.json, .env, provider databases, or shell profiles containing keys to Git.
Revoke and replace a key immediately if it may have been exposed.
Download updates only from the project's official repository or website.

Support boundaries#

BeeSwitch can help with your BeeSwitch account, API key, balance, model access, request logs, and gateway responses. Installation issues or behavior inside CC Switch, Codex++, or Cockpit Tools are controlled by those independent projects and may require help from their maintainers.
When contacting BeeSwitch support, include the approximate request time, model ID, tool name, operating system, and the BeeSwitch request ID if one is available. Do not send your API key.

Technical references#

CC Switch configuration reference
Codex CLI configuration reference
Claude Code configuration reference
OpenAI-compatible and Anthropic-compatible API reference
CC Switch source repository
Codex++ source repository
Cockpit Tools source repository
Modified at 2026-09-08 14:07:27
Next
Claude Code for VS Code Tutorial
Built with