Recently, a netizen discovered a more realistic problem while using WorkBuddy frequently:High-level models are very powerful, but in complex task scenarios, call costs and resource consumption will also increase rapidly.
This netizen recommended WorkBuddy many times before. In order to prepare for courses and corporate training, he began to use WorkBuddy intensively, especially in complex tasks, code analysis, scheme design, data collation and other scenarios, where high-capacity models almost became a necessity.
At the same time, a new problem gradually emerged:
If your computer already has subscription and login rights for other AI services, can you directly reuse these model capabilities instead of repeating configuration and payment?
For example, this user had originally subscribed to and logged in to Codex, Grok, and Google AntiGravity / Gemini on his computer. Now that these capabilities are already available, can they directly connect to WorkBuddy to form a unified AI work portal?
Based on this idea, the netizen developed a WorkBuddy CLI Model Bridge Skill。
It's based on open source projects CLIProxyAPI Build it, which can automatically detect locally logged in CLI Agents, complete model capability bridging, and allow WorkBuddy to directly use existing model resources such as GPT, Grok, and Gemini.
What is WorkBuddy CLI Model Bridge Skill?
Simply put, this is a bridge between WorkBuddy and the local CLI Agent.
It maps logged in CLI model capabilities such as GPT, Grok, and Gemini to WorkBuddy through the CLIProxyAPI.
Users do not need to:
- Manually find API Base URLs
- Apply and fill in the API Key yourself
- Modifying complex configuration files
- Study different model interface formats
Just let the Agent execute this Skill to complete the entire access process.
What problems does it solve?
Many AI users currently encounter similar problems:
1. Already subscribed to multiple AI services, but separated from each other
Many developers, researchers and heavy AI users also own:
- OpenAI Codex Subscription
- Grok permissions
- Gemini / Google AI Services
- Other CLI Agent tools
These services are already paid for, but are usually only available on your own clients.
Although WorkBuddy supports multiple model calls, many users still need to configure additional APIs.
The goal of this Skill is to open up these capabilities.
2. The cost of calling high-level models is increasing
The value of AI Agents lies in the continuous execution of complex tasks:
- long document analysis
- project planning
- code refactoring
- data processing
- automated workflow
If all these tasks use the highest-level model, the consumption will be very obvious.
Therefore, a more reasonable approach would be:
Use low-cost models for simple tasks;
High-capacity model for complex task invocation;
Existing subscription resources are reused first.
Skill's workflow
WorkBuddy CLI Model Bridge Skill automatically completes the following steps:
Step 1: Detect the local CLI Agent
It scans the current environment for installed and logged in AI CLI tools.
For example:
- Codex
- Grok CLI
- Gemini-related tools
Identify which model capabilities already exist.
Step 2: Automatically install or fix CLIProxyAPI
If there is no CLIProxy API in the environment, Skill will complete the installation automatically.
If there is an installation but the configuration is abnormal, it will also attempt to repair it.
Users do not need to understand the underlying proxy logic.
Step 3: Get a list of real models
Skill reads available models from the actual environment, rather than having users manually fill in the model name.
This can avoid:
- Wrong model name
- API version mismatch
- Configuration omission
Step 4: Automatically verify model capabilities
It's not just testing whether you can call it.
It will further verify:
- Ordinary text generation
- Tool calling capabilities
- Image processing capabilities
- reasoning ability
- Fast routing
Make sure the model written to WorkBuddy is truly usable.
Step 5: Automatically write WorkBuddy configuration
After the verification is passed, Skill will add the model to WorkBuddy.
At the same time:
- Keep existing manual configurations
- Does not overwrite the user's original settings
- Avoid disrupting existing workflows
Models that are currently running
Those currently passing the test include:
- GPT-5.6 SOL / Fast
- Grok 4.5
- Gemini 3.6 Flash
These models can already be called through WorkBuddy.
For daily office, technology development, content creation, corporate training and other scenarios, WorkBuddy's model capabilities can be significantly expanded.
Why choose CLIProxyAPI?
The core reason is that it provides a more flexible way to connect.
Traditional way:
用户 → API Key → 官方接口 → 模型
CLIProxyAPI provides:
本地 CLI Agent
↓
CLIProxyAPI
↓
WorkBuddy
↓
GPT / Grok / Gemini
This allows you to reuse existing login status and allow collaboration between different AI tools.
What does it mean for ordinary users?
The greatest value is to lower the threshold for use.
Previously:
To use GPT, Grok, and Gemini in WorkBuddy, you need:
- Registration API
- Create Key
- Check interface address
- modify the configuration
- Debugging model parameters
Now:
Just need:
- Install Skills
- Running Agent
- Wait for automatic configuration to complete
The complex work is left to AI itself.
The next stage of AI Agents: Not having more models, but connecting more models
In the past few years, everyone has focused on:
"Which model is the strongest?"
But as the number of models increases, new problems arise:
"How to make these models work together?"
Future AI Agents should not just be bound to a certain model, but should become an intelligent dispatch center:
- Select model based on task
- Choose routes based on cost
- Automatic switching based on capabilities
- Call user's existing resources
The WorkBuddy CLI Model Bridge Skill is just an attempt.
I hope it can help more users and truly connect the AI capabilities they already have.
If you are also using WorkBuddy, or if you have multiple AI Agents logged in to your computer, you might as well try this method to make the AI workflow truly realize "one call, multiple models collaborate."