Run Claude directly in your editor. No API calls. Full privacy. Maximum speed.
Inference runs on your machine. No network overhead, no API queues. Response times measured in milliseconds.
Your code stays on your machine. No telemetry, no logs sent anywhere. True offline-first operation.
One-time setup, unlimited usage. No per-request billing. Ideal for intensive development workflows.
Deep integration with your workspace. Understands your project structure, imports, and live code.
Add the extension from the VS Code marketplace. Select your local Claude model from available options.
Point to your local model, set inference parameters, and choose which project folders to index for context.
Open the assistant panel, ask questions about your code, generate completions, and refactor right in the editor.
| Feature | Local Extension | Claude API |
|---|---|---|
| Cost per query | Free | $0.003 - $0.015+ |
| Response latency | < 500ms | 2 - 10s |
| Data privacy | 100% local | Sent to Anthropic |
| Works offline | Yes | No |
| Setup time | 15 - 30 minutes | 2 minutes |
| Model updates | Manual | Automatic |
Four simple steps to run Claude locally in your editor:
Search for "VS Code Local LLM Assistant" in the Extensions marketplace and click Install.
The extension will guide you through downloading a Claude model quantized for local inference. Choose your preferred model size based on your hardware.
In VS Code settings, specify which workspace folders should be indexed for context-aware responses.
Open the assistant sidebar (Cmd+Shift+L) and begin asking questions. The model will be instantly available for code completion and project-wide Q&A.
No. Inference runs in a separate background process with its own CPU/GPU threads. VS Code remains responsive during model computation.
Yes. After the initial model download, the extension works completely offline. Useful for air-gapped environments or when traveling.
Quantized local models achieve 95%+ of the API model's accuracy with a 20-40% speed improvement and zero latency penalty. Minimal quality loss for massive speed gain.
Yes. The extension supports hot-swapping between different quantized models. Download once, use multiple times across projects.
Completely. All processing happens on your machine. Code never leaves your disk unless you explicitly commit it to git.
The extension supports variable quantization levels. Smaller models run on as little as 4GB RAM with acceptable latency trade-offs.
Join developers who've eliminated API latency and privacy concerns.
VS Code 1.70+ • Free & Open Source
Register interest
This is not a purchase and there is no card field. It puts your address, this product, and whatever you write below in front of a person, and you get a written answer about what finishing it, or handing it over for you to run yourself, would actually take.