Support custom model endpoint in Copilot CLI (like VS Code)
- Dominant language
- Shell
- Stars
- 11.2k
- Forks
- 1.9k
- Avg merge
- 14h 16m
- Merged PRs (30d)
- 6
Description
### Proposed solution
Title: Support custom model endpoint in Copilot CLI (like VS Code)
VS Code's Language Models panel allows configuring custom endpoints
for local or private models. Copilot CLI should have the same capability.
This would enable:
- Local model development and testing
- Enterprise private model deployment
- Integration with Ollama, LLaMA.cpp, etc.
Copilot CLI currently only supports GitHub's official models (Claude, GPT-5), but VS Code's Copilot extension supports custom model endpoints through the Language Models API.
I need to use a local model (running on my machine) or a custom API endpoint with Copilot CLI, just like I can in VS Code. This would enable:
1. **Local model development** - Test custom models locally before deployment
2. **Private/Enterprise models** - Use organization's private LLM endpoints
3. **Cost control** - Run models locally instead of using API quotas
4. **Integration with local tools** - Connect to Ollama, LLaMA.cpp, vLLM, or other local inference servers
## Proposed solution
Add support for custom model endpoints in Copilot CLI, similar to VS Code's Language Models panel:
```bash
copilot /model-config
# or
copilot --model-endpoint http://localhost:8000
```
Configuration in ~/.copilot/config.json:
```JSON
{
"customModels": [
{
"name": "my-local-model",
"endpoint": "http://localhost:8000/v1",
"apiKey": "optional-key"
}
]
}
```
Then select it in the /model picker just like official models.
### Example prompts or workflows
_No response_
### Additional context
VS Code already has this feature via Language Models API
Many users want to use local models for privacy/cost reasons
OpenAI-compatible API format is now standard (Ollama, vLLM, etc.)
This would make Copilot CLI feature-parity with VS Code in model selection
Contributor guide
Research direction
Start by reviewing the existing /model picker and configuration handling, then compare the proposed ~/.copilot/config.json customModels shape with the VS Code Language Models API described in the issue. Done means Copilot CLI can configure and select a local or private endpoint, including the OpenAI-compatible services named here such as Ollama, vLLM, or LLaMA.cpp.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- ollama, shell, vscode
- Domain
- ai, api, cli
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Quiet
- Clarity
- Mostly clear
- Newbie friendliness
- 38/100