Local AI Copilot Tools
English |
Español
DeepSeek V4 + OpenRouterAI + Ollama + LM Studio — all inside Copilot Chat's model picker.
Use DeepSeek V4, Open Router models, or local models (Ollama, LM Studio) directly from GitHub Copilot Chat — with agent mode, tool calling, vision proxy, and thinking mode.
Why this extension?
- Three providers, one picker. DeepSeek V4 Flash/Pro, Open Router (hundreds of models), and local backends (Ollama, LM Studio) — all in the familiar Copilot Chat dropdown.
- Don't replace Copilot — power it up. No new sidebar, no new chat UI. Just more models in the picker you already use.
- Agent mode, tool calling, instructions, MCP, skills — all of it still works. Copilot's entire stack, now running on your chosen model.
- Vision on text-only models. Drop an image into chat and the extension proxies it through a vision-capable Copilot model — transparently.
- BYOK. Each provider uses your own API key. Stored in the OS keychain, never on disk.
Features
DeepSeek V4 Pro & Flash
Both models with 1M token context, thinking mode with reasoning effort control, and transparent vision proxy. Switch models mid-chat without losing history.
Open Router
Access hundreds of models via Open Router's API. Configure which models to use — auto-detect or specify manually.
Local AI Backends (Ollama & LM Studio)
Run models locally on your machine. No API key needed — just install Ollama or LM Studio and point them to your models.
Transparent Vision Proxy
Text-only models can't see images. Drop a screenshot into chat — the extension hands the image to a vision-capable Copilot model, gets a description, and feeds it back. Zero config — just pick your preferred vision model once.
Thinking Mode
Full support for DeepSeek V4's reasoning_content. Use Copilot Chat's model picker menu to choose none, high, or max.
Inherits Every Copilot Capability
- Agent mode — autonomous multi-step tasks
- Tool calling — file edits, terminal, workspace search, Git, tests
- Instructions & skills — all your
.instructions.md, AGENTS.md, and skills
- Prompt caching stats — cache hit rate logged to the output channel
Secure by Default
API keys live in VS Code's SecretStorage (OS keychain). Never in settings.json.
Getting Started
Prerequisites
- VS Code 1.116 or later
- GitHub Copilot subscription (Free / Pro / Enterprise — the free tier works)
Usage
- Run Local AI: API Key DeepSeek from the Command Palette (
Cmd+Shift+P)
- Paste your DeepSeek API key from platform.deepseek.com
- Open Copilot Chat, click the model picker, pick any model
- Want Open Router? Run Local AI: API Key custom provider and add your Open Router key
- Have Ollama or LM Studio running? Models appear automatically
Models
| Provider |
Best For |
| DeepSeek V4 Flash |
Fast everyday coding, quick edits, cheap iteration |
| DeepSeek V4 Pro |
Complex refactors, agent tasks, deep reasoning |
| Open Router |
Hundreds of models via a single API |
| Ollama |
Local models (open source, private) |
| LM Studio |
Local models via local API server |
Settings
| Setting |
Default |
Description |
local-ai-copilot-tools.baseUrl |
https://api.deepseek.com |
DeepSeek API endpoint |
local-ai-copilot-tools.maxTokens |
0 |
Max output tokens (0 = no limit) |
local-ai-copilot-tools.modelIdOverrides |
prefilled official ID map |
API model IDs for DeepSeek V4 Flash / Pro |
local-ai-copilot-tools.visionModel |
(auto) |
VS Code vision model for image proxy |
local-ai-copilot-tools.visionPrompt |
(built-in) |
Prompt for image description |
local-ai-copilot-tools.experimental.stabilizeToolList |
false |
Stabilize tools parameter across turns |
localAiCopilotTools.openRouterModels |
(empty) |
Manual Open Router model list (comma-separated, empty = auto-detect) |
Available Commands
| Command |
Description |
Local AI: API Key DeepSeek |
Set/update/delete DeepSeek API key |
Local AI: API Key custom provider |
Set/update/delete Open Router API key |
Local AI: Activar/desactivar fuentes |
Enable/disable backends |
Local AI: Configurar visión proxy |
Configure vision proxy mode |
Local AI: Configurar modelos custom |
Configure Open Router models |
Local AI: Ver estado de conexión |
Show connection status dialog |
Local AI: Refrescar modelos |
Refresh available models |
Local AI: Test capacidades de modelo |
Test model capabilities |
License
Apache 2.0 — see LICENSE.
