Guide

Nock Coder in VS Code

Chat with Nock Coder in VS Code's own chat, on your own machine. It uses Ollama and the official Ollama extension. No account and no API key, and it keeps working with the Wi-Fi off. About five minutes.

01

Install Ollama

Windows shown here. On macOS and Linux, use the installer from ollama.com/download.

winget install Ollama.Ollama

02

Download Nock Coder

About one gigabyte, downloaded once.

ollama pull nockai/nock-coder

03

Install the Ollama extension

Needs VS Code 1.127 or newer. In VS Code, open Extensions, search for Ollama and install the one from the publisher Ollama. Or run:

code --install-extension Ollama.ollama

04

Pick Nock Coder in Chat

Open Chat (Ctrl Alt I, or Ctrl Cmd I on macOS). Open the model picker at the bottom of the chat input and choose nockai/nock-coder under Ollama. Local models need no sign-in.

Not in the list? Run Ollama: Refresh Models from the Command Palette, or reload VS Code. Ollama: Diagnose Models shows why a model is missing.

05

Use it

Ask: select some code, then ask about it in Chat. The selection is attached to your question.
Answers start with the code. Copy or insert it, then review it before you run it.

Want inline edits too?

Continue, a free and open source extension, adds edit a selection, apply and autocomplete. Continue was acquired by Cursor in 2026 and its last VS Code release was in June 2026. It still installs and works, but it may not get fixes.

01

Install Continue

Open Extensions, search for Continue and install the one from the publisher Continue. Or run:

code --install-extension Continue.continue

02

Point Continue at Nock Coder

Open the Continue panel, click the config selector at the top and open the local config for editing (usually ~/.continue/config.yaml). Replace its contents with this and save. The model name must match ollama list exactly.

name: Nock Coder
version: 0.0.1
schema: v1
models:
  - name: Nock Coder
    provider: ollama
    model: nockai/nock-coder:latest
    roles:
      - chat
      - edit
      - apply
  # Optional: autocomplete (uses a base model)
  - name: Autocomplete
    provider: ollama
    model: qwen2.5-coder:1.5b-base
    roles:
      - autocomplete

The second model is optional. Nock Coder is tuned for answering questions, not for completing code as you type, so autocomplete uses Qwen2.5 Coder base. Get it with ollama pull qwen2.5-coder:1.5b-base, or delete those lines.

03

Use it

Ask: select some code, press Ctrl L (Cmd L on macOS) and ask a question about it.
Edit: select code, press Ctrl I (Cmd I) and describe the change. Accept or reject the diff.
Pick Nock Coder in the model menu at the bottom of the chat if another model is selected.

What it is good for

Good at: Short snippets, explaining or fixing one function, small edits to a selection, quick questions while you work. Answers start with the code.
Not for: Agent mode, which needs a model that calls tools, and large changes across many files. It is a 1.5B model, so expect mistakes and review every change.
Privacy: Everything runs on your machine through Ollama. Your code is not sent to NockAI or anyone else.
Limits in VS Code: VS Code's own chat uses local models for chat only. Inline suggestions as you type still need a GitHub account and do not use local models.