Use Ollama in Chrome, right in the text box you are typing in
ShroomPen is a free Chrome extension that sends your writing requests to the Ollama server on your own computer and puts the answer into the text field you are using, such as a webmail reply, a CMS editor or a LinkedIn post. You point it at http://localhost:11434/v1, allow the extension in OLLAMA_ORIGINS once, and pick a model. No API key, no ShroomPen account, and with a local model the request never leaves your machine.
Published
What you get
Most Ollama browser extensions give you a chat window next to the page. You ask, then copy the answer into the field you were working in. ShroomPen works the other way round. You click into the field first, open ShroomPen over the page, and the draft goes back into that field when you click Insert draft. If you selected text, it can replace the selection instead.
- Reply to an email in webmail such as Gmail, with the thread as context.
- Rewrite a paragraph in a CMS, a support tool or a web form.
- Draft a LinkedIn post or comment, or translate what you wrote.
You choose the context for each request: the page, the field you are in, the text you selected, other open tabs (PDF tabs too) or a saved Workspace. Password, one-time code and payment fields are left out.
Set it up
Install Ollama and pull a model
Install Ollama from ollama.com (Windows 10 or later, macOS 14 or later, or Linux), then download a model, for example with ollama pull llama3.2 (about 2 GB).
ollama pull llama3.2Allow ShroomPen in OLLAMA_ORIGINS
Ollama rejects requests from browser extensions until you allow them. Set the environment variable OLLAMA_ORIGINS to ShroomPen’s origin, shown below, and restart Ollama.
OLLAMA_ORIGINS=chrome-extension://ghigbnpoahodgnhlgbggjphpflkofgfnWhere to set it depends on your system. The OLLAMA_ORIGINS guide has the steps for Windows, macOS and Linux and a one-line check that it worked.
Install ShroomPen and enable it
Add ShroomPen from the Chrome Web Store. The first time you open it, it explains what it can read and waits until you tick the box and click Enable ShroomPen.
Add ShroomPen to Chrome. It is free. You can also open it from the toolbar button or by right-clicking a page and choosing Open ShroomPen.
Add Ollama as a local server
Open ShroomPen with Ctrl+Q, open settings, choose Add provider, then Direct provider, then Local server. Enter http://localhost:11434/v1 as the base URL and leave the API key empty.
http://localhost:11434/v1Test and connect
Click Test/load models and allow access to localhost in the permission window Chrome shows. ShroomPen loads your Ollama models and sends a short test request. Pick a model and click Connect.
Write in a text box
Click into any text field on a page, press Ctrl+Q, type what you want, read the draft and click Insert draft. ShroomPen never changes a field on its own.
Where your text goes
ShroomPen has no server of its own. The request goes from Chrome straight to the address you entered, here Ollama on localhost. Your provider settings are kept in the extension’s local storage in your Chrome profile. Recent conversations are kept only until you close the browser, and the page text you attach is not saved unless you add it to a Workspace yourself. You can open the exact request before it is sent.
One exception to “it stays on your machine”: Ollama can also run cloud models, which run on Ollama’s servers after you sign in to Ollama. If you pick one of those in ShroomPen, your request goes to Ollama’s cloud. Local models don’t.
ShroomPen or another Ollama extension?
The best-known alternative is Page Assist, an open-source sidebar and web UI for local models. They solve different problems, and Page Assist is ahead on several points.
| ShroomPen | Page Assist | |
|---|---|---|
| Main use | Write into the text field you are using | Chat in a sidebar or a full-page web UI, chat with the page |
| Result goes into the page | Yes, Insert draft or replace the selection | You copy it yourself |
| Needs OLLAMA_ORIGINS | Yes, once | Usually not: it rewrites request headers for localhost addresses |
| Browsers | Chrome and Chromium browsers that install from the Chrome Web Store | Chrome, Edge, Brave and Firefox |
| Source code | Closed | Open source |
| Hosted providers | OpenRouter (one-click sign-in), OpenAI, Anthropic, Gemini, any OpenAI-compatible API | Many OpenAI-compatible providers, added with an API key |
If you mostly want to chat with a local model, or you need Firefox or open source, Page Assist is the better fit. If you want the model’s draft in the email or form you are writing, that is what ShroomPen is for. Both can talk to the same Ollama server.
Related
- Set OLLAMA_ORIGINS for a Chrome extension: the 403 fix in detail
- Use OpenRouter in Chrome: hosted models, including free ones, with one sign-in
- Privacy policy: what ShroomPen handles and stores
Questions
Is there an Ollama Chrome extension that writes into web pages?
Yes. ShroomPen connects to Ollama through its OpenAI-compatible API at http://localhost:11434/v1 and inserts the result into the text field you are using. Most other Ollama extensions are sidebar or tab chats where you copy the answer yourself.
Why does ShroomPen get a 403 error from Ollama?
Ollama only accepts browser requests from origins on its allow list, and chrome-extension:// origins are not on it by default. Set OLLAMA_ORIGINS to chrome-extension://ghigbnpoahodgnhlgbggjphpflkofgfn and restart Ollama.
Does ShroomPen need an API key for Ollama?
No. Choose Local server and leave the API key field empty. Ollama ignores keys on its local API.
Does my text leave my computer when I use Ollama?
Not with a local model. ShroomPen has no server of its own, so the request goes from Chrome to Ollama on localhost. Ollama cloud models are different: Ollama runs those on its own servers.
Can I use Ollama running on another computer?
Only over HTTPS. ShroomPen allows plain HTTP for localhost, 127.0.0.1 and ::1 only, so a server elsewhere on your network needs an HTTPS address, for example through a reverse proxy.
Which Ollama model should I use for writing?
Start with a small model such as llama3.2 (3B) or gemma3:4b if you have no dedicated GPU, and try a larger one if the drafts are too plain. ShroomPen waits up to 45 seconds for an answer, so very large models on a CPU can time out.
Does it work with LM Studio too?
Yes. Choose Local server and use LM Studio’s address, which is http://localhost:1234/v1 by default.