Recoll on Windows: What It Is and How to Set Up Full-Text Search

Recoll is a local, open-source full-text search tool built on the Xapian search engine. On Windows it runs as a native application and indexes the contents of your files—not just filenames—so you can find a document by a phrase inside it. It fits best if you want a free, configurable, offline indexer and are comfortable adjusting settings; if you want a zero-configuration tool that parses Office, PDF, and image text out of the box, a purpose-built Windows searcher may suit you better.

What Recoll actually does

Recoll walks the folders you point it at, extracts text from each file using external helper programs ("filters"), and stores the result in a Xapian index. Queries run against that index, so searches return in well under a second even across large collections.

Two consequences follow from this design:

• Coverage depends on filters. Recoll itself does not parse every format. It calls tools such as pdftotext (Poppler) for PDFs, antiword or catdoc for older Word files, and unzip-based helpers for OOXML. If a filter is missing, the file is indexed by name only.
• The index is a snapshot. New and changed files are picked up on the next indexing run, not instantly, unless you schedule or trigger it.

Installing Recoll on Windows

1. Get the installer. Download the Windows build from the official Recoll site (lesbonscomptes.com/recoll). The Windows package bundles the required runtime, so you generally do not need to install Python or Xapian separately.
2. Run the installer and complete the setup wizard. Accept the default install path unless you have a reason to change it.
3. Launch Recoll. On first start it opens the Index Configuration dialog, because no index exists yet.

If the installer reports a missing dependency, install that component and re-run setup. The bundled build is the path of least resistance; compiling from source on Windows is possible but rarely worth it.

Configuring what gets indexed

In the Index Configuration dialog:

• Top directories — add the folders to index (for example C:\Users\You\Documents). Add each root separately; Recoll recurses into subfolders.
• Skipped paths — exclude noise such as node_modules, .git, AppData\Local\Temp, and build output directories. This is the single biggest lever on index size and speed.
• File types — leave the defaults to index everything with a known filter, or restrict to specific extensions if you only care about documents.
• Start indexing. The first run on a large drive can take a while; subsequent runs only process changes.

Keeping the index current

Recoll does not watch the filesystem continuously by default. To keep results fresh you either:

• run Index → Update Index manually, or
• schedule recollindex (the command-line indexer) with Windows Task Scheduler to run on a cadence that matches how often your files change.

Searching

Type a query in the main window. Useful syntax:

Results show a snippet with the matched terms highlighted. Double-click to open the file in its default application, or use the preview pane to check context before opening. You can sort by relevance, date, or path, and filter by file type or directory from the results view.

Common problems and fixes

• New files don't appear. The index hasn't been updated. Run an update or check that your scheduled recollindex task is actually firing.
• PDFs or Office files return no content hits. The relevant filter is missing or failed. Confirm the helper tool is installed and on PATH, then re-index those files.
• Non-English text isn't found. Indexing and query stemming are language-dependent. Set the correct language in preferences so tokenization matches your content.
• Permission errors during indexing. Recoll can't read protected folders. Remove them from the top directories or run the indexer with sufficient rights.
• Index is huge or slow. Tighten skipped paths and file-type filters; a bloated index is almost always caused by indexing caches and dependency folders.

How Recoll compares on Windows

Anytxt's own documentation positions it as a full-text engine that extracts text from Office, PDF (including scanned PDFs via OCR), eBooks, and images without installing extra software, and describes it as a Google Desktop Search alternative. That built-in parsing is the main practical difference from Recoll, where each format depends on an external filter you must supply.

Choosing between them

Pick Recoll if you value an open-source, scriptable indexer and don't mind configuring filters and a scheduled update. Pick Anytxt if you want content search across many formats—especially scanned PDFs and images—without assembling dependencies. Pick Everything if filename search is all you need and speed is the priority.