OllaCoder

OllaCoder is a private AI coding assistant for VS Code. It runs local models through Ollama, so your code never leaves your machine.

Features

Chat with an AI about your code, with full streaming
Agent mode — reads, edits files and runs commands under your control
Inline edit (Ctrl+I) — select code, describe the change, accept the diff
Autocomplete — Tab-completion while you type, using a dedicated FIM model
@codebase — local vector index of your project for grounded answers
Custom prompts, modes and tools — fully configurable from the settings panel
Optional cloud providers (OpenAI, Groq, OpenRouter, DeepSeek, Mistral, …) when you need them

Requirements

VS Code 1.85 or newer
A working Ollama install with at least one model pulled, e.g.:

ollama pull qwen3:8b              # chat / agent
ollama pull qwen2.5-coder:1.5b    # autocomplete
ollama pull nomic-embed-text      # @codebase embeddings

Getting started

Install the extension
Make sure Ollama is running
Open the OllaCoder sidebar (Ctrl+Shift+L)
Pick a model and start chatting

Configuration

The most common settings — the full list is under OllaCoder in VS Code settings.

Setting	Default	Notes
`ollacoder.ollama.baseUrl`	`http://localhost:11434`	Ollama API URL
`ollacoder.models.chat`	`qwen3:8b`	Model used for chat & agent
`ollacoder.models.autocomplete`	`qwen2.5-coder:1.5b-base`	FIM model for Tab-completion
`ollacoder.models.embeddings`	`nomic-embed-text:latest`	Used by `@codebase`
`ollacoder.agent.maxIterations`	`25`	Cap on agent reasoning steps

License

This extension is not open source. The source code, design, and assets are owned by the author. You may install and use the extension; you may not copy, modify, redistribute, reverse-engineer or republish any part of it without prior written permission.