Local models · free

Write with your own model — free

Run a language model on your own computer through Ollama and use it right inside the studio. The compute is yours, so no Ink is spent and your text never leaves your machine.

Free

The model runs on your computer. Generating pages spends no Ink.

Private

Text is processed locally. No cloud provider ever sees it.

No port forwarding

The bridge makes an outbound connection to the service — it works behind routers and firewalls.

What you need

A computer that runs the model comfortably (for prose, 8 GB+ free memory helps).
Ollama — a free app that runs models locally.
Our small bridge app — one download, nothing else to install.

Supported systems

macOS — Apple Silicon (arm64) and Intel (x86_64)
Windows — x64 and ARM64
Linux — x64 and ARM64

Step-by-step

Pick your operating system — the commands adapt.

Step 1. Install Ollama and pull a model

Ollama is the engine that runs the model on your computer. After installing, pull a model with “ollama pull <name>”.
macOS · terminal
```
brew install ollama
ollama pull llama3.2
```
Examples: llama3.2 (small and fast, ~2 GB) for a first try; qwen2.5:7b or mistral (better prose, ~5 GB) if you have the memory. Any Ollama model works.
Step 2. Get a bridge token

In the studio, click the model name at the top to open the picker, then “Connect machine” → “Create a token”. The token is shown only once — copy it.

Step 3. Download and run the bridge

Paste your token, then copy this one command — it downloads the bridge for your system and starts it. Keep the window open: while it runs, your model is available in the service.

Your bridge token Stays in your browser — only used to build the command below.

Copy and run this command:

macOS · terminal

ARCH=$([ "$(uname -m)" = arm64 ] && echo aarch64 || echo x86_64)
curl -fsSL "https://github.com/fedorello/penfreely-bridge/releases/latest/download/penfreely-bridge-${ARCH}-apple-darwin.tar.gz" | tar xz
xattr -d com.apple.quarantine penfreely-bridge 2>/dev/null; chmod +x penfreely-bridge
PENFREELY_BRIDGE_TOKEN="<your-token>" \
  PENFREELY_BACKEND_WS_URL="wss://api.penfreely.com/bridge/connect" \
  ./penfreely-bridge

Prefer to download manually? All builds →

Step 4. Select your model in PenFreely and write

With the bridge running (step 3), go back to the studio and:
1. Click the model name at the top of the studio to open the model picker.
2. Find the “Your machine · local” group — your local models are listed there, each with a green dot and a “free” tag.
3. Click the model you want. Then write or generate a page as usual — the text streams from your own model and no Ink is spent.
How the picker looks while the bridge is running.

The green dot means the model is online. If you close the bridge app, the model greys out — start it again to bring it back.

Common questions

Is it really free?

Yes. The model runs on your computer, so generating with a local model spends no Ink — on any plan, including Free.

I don’t see my model in the picker

Make sure the bridge app is running and the model is pulled in Ollama (ollama list), then reopen the picker. Local models appear only while the bridge is connected, under “Your machine · local”.

Do I need to install anything else?

No. Just download the one small bridge app and run it. Nothing else is required.

Which model should I pick?

Any model available in Ollama. Small ones (1–3B) are fast but plainer; 7–14B models give noticeably better prose if you have the memory.

What if I turn my computer off?

The model goes offline and is shown as unavailable. Just run the bridge again — it reconnects automatically.

Is it safe?

The connection is outbound only; no inbound ports are opened. The token is shown once, stored hashed, and revocable any time in the studio.

Ready to write with your own model?

Open the studio, get a token, and connect your computer in a couple of minutes.

Open the studio