Built for Obsidian · public beta

Find the note
you meant.

ScottSearch spends a moment understanding your question, then returns ten useful notes—not a hundred places where the same word appears.

Free and open source. Your notes stay local by default.

ScottSearch 10 results
reducing coordination overhead
Meaning + wording 0.8 seconds
When meetings are worth itNotes/Collaboration
88

A shared document handles status; reserve live time for disagreement and discovery…

Project handoff checklistTemplates
79

10carefully ranked results

0accounts required

100%open source

Why another search?

Less hunting.
More finding.

Regular search is excellent at answering “where did I type this word?” ScottSearch is for “where was that idea?”

01

Meaning before volume

Optional local AI compares ideas, while a strong wording score keeps precise matches grounded.

02

Power when you want it

Require a phrase, exclude a term, or filter by tag, path, and dates without learning a complicated language.

+"exact phrase" -draft
03

Built into Obsidian

Open it from the ribbon or command menu, move with arrow keys, and keep the theme you already chose.

Enter

Try the human-testing release

Install without touching code.

ScottSearch 0.1 uses the same small plugin bundle on macOS, Windows, Linux, iPhone, iPad, and Android. During beta, BRAT is the easiest installer.

One downloadAll platforms

Download the ZIP

The ZIP contains the three plugin files Obsidian needs; the repository also carries readable license information. This route is useful when BRAT is not an option, but it requires access to your vault's hidden plugin folder.

Human-testing release0.1.3
Pre-release
Download ScottSearch 0.1.3
Show manual installation steps
  1. Unzip the download.
  2. Create a folder named scottsearch inside <your vault>/.obsidian/plugins/.
  3. Move main.js, manifest.json, and styles.css into that folder.
  4. Restart Obsidian. Open Settings → Community plugins and turn on ScottSearch.

On phones and tablets, hidden folders are awkward to manage. BRAT is usually the friendlier option.

A roadmap you can read

Where ScottSearch is going.

The simple view is here. Every card links to the detailed public work item if you want the full story.

Available now

Search foundation

Meaning-aware ranking, power queries, filters, keyboard control, and title-only, default, or verbose result detail.

Measured relevance done
02

Human testing

Real-vault feedback

We are learning which results feel useful, how much preview context helps, and what breaks across devices.

Beta program active
03

After testing

Community directory

Automated Community checks are ready. Next: resolve beta findings, publish a stable build, and complete Obsidian's review.

Official submission planned
04

Prototype in progress

Zero-setup semantics

The consent-first download and desktop worker prototype now work. A public build waits for macOS, Windows, and Linux Obsidian measurements; mobile stays off pending its own tests.

Desktop experiment prototype
Open the detailed public roadmap

No technical background needed

Five useful things to test.

There is no wrong feedback. A sentence about what you expected and what happened is enough.

  1. 1

    Find a remembered idea

    Describe a note without copying its exact words. Is a useful note in the first ten?

  2. 2

    Be very specific

    Try +"an exact phrase", -draft, or a date such as createdafter:2025-01-01.

  3. 3

    Sort, narrow, and choose detail

    Try the filters, then cycle Title only, Default, and Verbose. The same results should update immediately.

  4. 4

    Use your keyboard

    Move with arrow keys, open with Enter, open a new tab with modifier-Enter, and clear with Escape.

  5. 5

    Try a failure

    If you use Ollama, stop it briefly. ScottSearch should explain the problem and keep searching by wording.

A successful test is useful too: the short beta form records which platform, theme, vault-size range, and search mode worked. Approximate answers are welcome, and you can choose “Not measured” for timings.

Keep private notes private: make up a fictional note and query that reproduce the same behavior before sharing feedback publicly.

Help shape it

You do not need to code.

Good software starts with a clear story about a real problem. Choose the contribution that feels comfortable.

Share a need

Our friendly form asks what is difficult today, what a better outcome looks like, and how we would know it is solved.

Open the requirement form

Test the beta

Try the five prompts above and report only the moment where reality differed from what you expected.

Use the test guide

Contribute code

Developers can pick a planned issue, run the documented checks, and propose a focused change.

Read the developer guide

Latest download

ScottSearch 0.1.3

The result-detail patch: Verbose now reliably expands beyond the Default preview without repeating the search.

Prepared September 2026

Download the ZIP All releases and individual files Static release information shown

Questions are welcome

Plain answers.

Do I need AI or Ollama?

No. ScottSearch includes a strong local wording-based search. Ollama is optional and adds matches based on meaning, including notes that use different words.

Do my notes leave my device?

Wording-based search stays inside Obsidian. Semantic search is off until you enable it. In 0.1, text goes only to the Ollama address shown in settings; the default is your own computer. An unreleased desktop experiment instead processes text inside an isolated on-device worker.

Which download is right for my computer or phone?

There is only one ScottSearch plugin bundle for every platform. BRAT is the easiest beta installer, especially on mobile. Semantic mode additionally needs an Ollama endpoint that the device can reach.

How do I remove it?

In Obsidian, open Settings → Community plugins, turn off ScottSearch, and choose uninstall. If you used BRAT, you can also remove ScottSearch from BRAT's plugin list. Your notes are not deleted.

Why is this called a human-testing release?

The main behaviors are built and automatically tested, but the plugin still needs feedback across real vault sizes, themes, and devices before official community-directory submission.

Search should end with recognition, not exhaustion.

Find the note you meant.

Try ScottSearch View on GitHub