Quick Start

Get ZenCoder running in under 5 minutes. Connect a local model and send your first chat.

Pre-install: macOS setup

Install the tools ZenCoder uses as its AI backends.

1. Install Ollama (local AI, free)

1. 1

   Download Ollama

   Go to ollama.com and download the macOS app, or use Homebrew:

   brew install ollama

2. 2

   Start the Ollama service

   Ollama runs as a background service on localhost:11434.

   Terminal 1bash

   ollama serve

   ℹ️

   On macOS, the Ollama app starts this automatically after install.
3. 3

   Pull a coding model

   We recommend one of these for coding tasks:

   # Recommended: fast, great for code
   ollama pull qwen2.5-coder:7b
   
   # Alternative: general purpose
   ollama pull llama3
   
   # Verify it downloaded
   ollama list

   Expected output: model appears in the list with its size.

4. 4

   Verify Ollama is running

   curl http://localhost:11434/api/tags
   # → {"models":[{"name":"qwen2.5-coder:7b",...}]}

2. Install LM Studio (optional alternative)

LM Studio is a GUI app that runs local models and exposes an OpenAI-compatible API.

1. 1

   Download LM Studio

   Download from lmstudio.ai — macOS, Windows, and Linux supported. Install the .dmg as you would any Mac app.

2. 2

   Download a model

   Open LM Studio → search for Qwen2.5-Coder or Mistral 7B → click Download. Models are stored in ~/.cache/lm-studio/models/.

3. 3

   Start the local server

   In LM Studio: Local Server tab → select model → Start Server.

   The server listens on http://localhost:1234/v1 (OpenAI-compatible).

Install ZenCoder

Choose how you want to use ZenCoder — the CLI tools or the VS Code chat extension (or both).

Option 1 — CLI & zencoder-secrets

Installs the zencoder CLI and zencoder-secrets in one step. Works on macOS, Linux, and Windows WSL2.

curl -fsSL https://raw.githubusercontent.com/divyabairavarasu/zencoder-releases/main/install.sh | bash

1. 1

   Open the REPL

   zencoder

   You'll see the ZenCoder banner and a > prompt. Type a question and press Enter twice to send.

   > What does this code do?
   |
   ZenCoder: This code defines an HTTP server that...

2. 2

   Check health

   zencoder health
   # → ZenCoder: service status is ok. Model: ollama/qwen2.5-coder:7b

Option 2 — VS Code Chat Extension

Get AI chat, inline completions, and context-aware suggestions directly inside VS Code.

Add cloud API keys with zencoder-secrets (optional)

zencoder-secrets is installed automatically alongside the CLI. It is the most secure way to add your own cloud API keys — every key is stored encrypted on your local machine and never leaves your device.

How it works

You register a provider endpoint and model once. ZenCoder handles routing — you just pick the model in the REPL.

# General syntax
zencoder-secrets add -url <provider-base-url> -model <model-id> -alias <your-key-name>
# → Enter API key: [hidden — input is masked]
# → BYOK provider registered.

NVIDIA NIM — featured example

NVIDIA NIM provides free API access to top-tier models including Llama, Mistral, and Nemotron. Sign up at build.nvidia.com to get a free key (no credit card required).

# Add an NVIDIA NIM key
zencoder-secrets add -url https://integrate.api.nvidia.com/v1 \
  -model nemotron-3-nano-30b-a3b \
  -alias nvidia-key1
# → Enter API key for nvidia (alias: nvidia-key1): [hidden]
# → BYOK provider registered.
# → Example model: nvidia/nemotron-3-nano-30b-a3b

# Add more keys for different models (ZenCoder load-balances across them)
zencoder-secrets add -url https://integrate.api.nvidia.com/v1 \
  -model llama-3.3-70b-instruct \
  -alias nvidia-key2

# Verify your keys
zencoder-secrets list

# Test a key is working
zencoder-secrets test -provider nvidia -alias nvidia-key1

Managing keys

Essential slash commands