InvAit for Visual Studio Code

Local First: No code leaves your machine by default.
Private AI: Native support for local LLMs (Ollama, LM Studio, vLLM).
Control: You define the endpoint and API keys. No telemetry.

Local AI coding agent with tool-calling support, running entirely inside VS Code via an embedded Blazor WebAssembly UI.

🔐 Security & Privacy

AI Chat — streaming chat with OpenAI-compatible LLMs (SSE), session history, and multi-model support.
Agent Tools — read/write/edit files, grep, search, run bash commands, git operations — all from the chat.
Context Awareness — automatically publishes active file content, selection, and workspace structure to the LLM.
Skills & Rules — load reusable skill files (*SKILL.md) and project/global rules (.agents/rules.md).
MCP Support — connect to external MCP servers (stdio/HTTP-SSE) for additional tools.
Sub-Agents — delegate complex subtasks to focused agent instances with their own tool budgets.
Approval Modes — per-category tool approval (Allow / Ask / Deny) for safe autonomous operation.

Server	Run description	Endpoint
unsloth	install / download model / create api token	http://localhost:8888
LM Studio	install / download model / start local server	http://localhost:1234
Olama	install / download model	http://localhost:11434

Server	Run description	Endpoint
OmniRoute	in terminal install `npm install -g omniroute` run `omniroute`	http://localhost:20128
OpenRouter	sing in / pay / use	https://openrouter.ai/api
Siliconflow	sing in / pay / use	https://api.siliconflow.com

/v1 at the end of the endpoint does not need to be specified. It will be automatically deleted and that's normal.

General	Tools

MIT

Run the command "Open InvAit chat" from the Command Palette (Ctrl+Shift+P / Cmd+Shift+P).
Configure your API endpoint and model in the settings dialog.
Start chatting - the AI can read files, edit code, run commands, and more.