Unlock Full‑Stack AI Coding on Ubuntu with Ollama and CC Switch

Many AI coding assistants such as Claude, Codex, and Gemini rely on cloud APIs, which introduce latency and privacy concerns. The article demonstrates how to replace these services with a fully local, free AI programming environment on Ubuntu using Ollama and CC Switch.

Step 1 – Install Ollama and prepare models

Install Ollama with the official script: curl -fsSL https://ollama.com/install.sh | sh Configure Ollama to listen on all interfaces by editing its systemd service and adding the environment variables:

[Service]
Environment="OLLAMA_HOST=0.0.0.0:11434"
Environment="OLLAMA_ORIGINS=*"

Reload and restart the service:

sudo systemctl daemon-reload
sudo systemctl restart ollama

Download and run a coding‑oriented model, e.g., Qwen2.5‑Coder 32B (requires ≥24 GB VRAM) or DeepSeek‑R1:

# Qwen2.5‑Coder 32B
ollama run qwen2.5-coder:32b
# or DeepSeek‑R1
ollama run deepseek-r1:14b

Step 2 – Install AI CLI clients

Install the command‑line interfaces that normally talk to remote APIs:

npm install -g @anthropic-ai/claude-code
npm install -g @openai/codex
npm install -g gemini-chat-cli

These commands would normally prompt for an API key; the next step will intercept the traffic.

Step 3 – Install and configure CC Switch

CC Switch (Claude Code Switch) is an open‑source proxy router. Install it via npm (or download the Linux binary): npm install -g @songhe/cc-switch Create a new provider configuration that points to the local Ollama service:

# Add a new profile named local-ollama
ccs new local-ollama

During the interactive setup, select:

Provider Type: OpenAI Compatible Base URL: http://localhost:11434/v1 API Key: ollama (placeholder)

Model: e.g., qwen2.5-coder:32b Activate the profile and start the proxy:

ccs switch local-ollama
ccs proxy start

CC Switch now forwards Claude, Codex, or Gemini requests to the local Ollama backend.

Final test – Run Claude locally

Execute the Claude CLI; it connects to the local Ollama instance without requiring an Anthropic account: claude Enter a prompt such as >> help me write a Snake game in Python. The Ollama engine performs the inference and the Claude CLI prints the generated code within seconds.

Advanced tips

Mix‑and‑match profiles: switch quickly between a local DeepSeek profile and a cloud‑based Claude‑3.5 profile for sensitive vs. heavy‑weight tasks.

Hot model switching: modify the model field in the CC Switch configuration to change the backend model on the fly.

By combining Ubuntu, Ollama, and CC Switch, developers obtain the interactive experience of commercial AI assistants while keeping all data on‑premise and eliminating costly API subscriptions.