Claude Code Local: Run Claude Code Free With Local Models

Claude Code Local is the open-source project that finally lets you run Claude Code completely free with local models.

Gemma 4, Qwen 3.5, Llama 3.3 — all supported.

Run offline.

Run without an API key.

Run without a subscription.

And the experience is properly good.

Let me show you how to set it up.

Video notes + links to the tools 👉

What Is Claude Code Local?

Claude Code Local is an open-source project that makes Claude Code work with local AI models.

Normally, Claude Code uses Anthropic's Claude models (paid API).

Claude Code Local swaps that out for free local models via Ollama.

Why this matters:

£0/month instead of subscription costs
Offline capable — no internet needed
Privacy — nothing leaves your machine
Unlimited usage — no rate limits

This is genuinely one of the most useful AI tool releases of 2026.

Supported Local Models

Claude Code Local works with any Ollama-compatible model:

Best Options

Gemma 4 — fast, lightweight, good quality
Qwen 3.5 — most capable, best quality
Llama 3.3 — reliable, well-tested

My Recommendation

Qwen 3.5 for best output quality.

Gemma 4 for fastest local performance.

Llama 3.3 as a balanced middle ground.

You can switch between models any time.

Quick Start Setup

Step 1: Install Ollama

# macOS
curl -fsSL https://ollama.com/install.sh | sh

# Or download from ollama.com

Step 2: Download a Model

ollama pull qwen3.5
# or
ollama pull gemma4
# or
ollama pull llama3.3

Step 3: Install Claude Code Local

Copy the quickstart commands from the project's GitHub.

Paste into your terminal.

Run.

Step 4: Launch

Run the Claude Code Local command.

Select your model.

Start using Claude Code locally.

If you want the basics of Ollama first, my Ollama + Hermes setup guide is a gentle introduction.

Why Run Claude Code Locally?

1. Cost Savings

Claude Code subscription costs £20-200/month.

Claude Code Local: £0 forever.

Over a year: £240-£2,400 saved.

2. Privacy

Your code never leaves your machine.

Critical for:

Proprietary codebases
Client work under NDA
Regulated industries
Personal projects

3. Offline Work

Travelling without internet?

Still have full Claude Code functionality.

4. No Rate Limits

Managed Claude Code has token limits.

Local has whatever your hardware can deliver.

Unlimited usage for the cost of electricity.

🔥 Want the complete Claude Code Local tutorial?

Inside the AI Profit Boardroom, I've added a full setup video with step-by-step configuration, model comparisons, and performance optimisation. Plus configurations for the best models. 3,000+ members running Claude Code Local productively.

→ Get the full Claude Code Local training here

Performance Comparison

Claude Code (Cloud)

Speed: Very fast
Quality: Highest
Cost: £20-200/month
Offline: No

Claude Code Local + Qwen 3.5

Speed: Moderate (depends on hardware)
Quality: Very good (80-90% of Claude Opus)
Cost: £0
Offline: Yes

Claude Code Local + Gemma 4

Speed: Fast
Quality: Good for simpler tasks
Cost: £0
Offline: Yes

For most day-to-day work, local models are surprisingly capable.

For complex long-context work, cloud Claude is still better.

Hardware Requirements

Minimum

16GB RAM
Any modern CPU (M1/M2/M3 Mac ideal)
50GB disk space per model

Apple Silicon Optimisation

M-series Macs run Ollama particularly well.

Unified memory architecture helps with larger models.

M4 Max specifically excellent for Qwen 3.5.

Use Cases for Claude Code Local

Private Client Work

NDA-protected projects.

Local inference keeps everything confidential.

Travel/Flight Work

No internet required.

Full functionality preserved.

Cost-Sensitive Operations

High-volume coding tasks.

Cloud would be prohibitively expensive.

Experimentation

Try approaches without burning API credits.

Unlimited iteration for free.

Learn how I make these videos 👉

Combining Local and Cloud

Best setup: both options available.

Switch based on the task:

Simple tasks → local (free, fast)
Complex reasoning → cloud (better quality)
Private code → local (privacy)
Time-sensitive → cloud (faster)

You get maximum flexibility + minimum cost.

Claude Code Local Speaking Feature

One cool feature — Claude Code Local can speak to you.

Text-to-speech integration lets you have your code read back.

Useful for:

Accessibility
Listening while working
Code review walkthroughs
Learning unfamiliar codebases

🔥 Master Claude Code Local for your workflow

Inside the AI Profit Boardroom, I share performance-optimised configurations for different hardware. Model selection strategies, quantisation choices, context management. Get maximum capability from your local setup.

→ Get the optimisation training here

Integrating With Other Tools

Claude Code Local works with:

Your existing codebase (any language/framework)
Git workflows (same as cloud Claude Code)
Ollama ecosystem (all Ollama models)
Your IDE (VS Code, JetBrains, etc.)

For AI agent context, see my Claude Code AI SEO breakdown on using Claude Code for content automation.

Claude Code Local: Frequently Asked Questions

Does Claude Code Local have the same features as the cloud version?

Most core features work. Some features tied to cloud services may be limited.

Which model should I pick for coding specifically?

Qwen 3.5 for general coding. Gemma 4 if speed matters more than quality.

Can I run Claude Code Local on Windows?

Yes, via Ollama Windows support. Performance varies with hardware.

How does offline Claude Code Local work?

Ollama runs models locally. Claude Code Local interfaces with Ollama. Everything on your machine.

Can I switch between local and cloud within one project?

Yes. Use local for some tasks, cloud for others. No conflict.

Is the code secure with Claude Code Local?

More secure than cloud — nothing leaves your machine. Good for sensitive work.