A Private Local Search Tool for Small Law Firms

October 20, 2026 · 7 min read

Most small firms don't lose documents. They lose the path to them. The clause you drafted for one client three years ago, the expert letter in a matter you've since closed, the scanned settlement agreement someone saved as "scan_0042.pdf". It's all on the firm's drives, but the only way to find it is to remember the matter, the folder and the filename.

Large firms solve this with a document management system (DMS), often cloud-hosted. Small and solo practices frequently don't want the cost, the migration or the third-party data exposure. This article explains what a local search tool for small law firms should do, how it compares with a cloud DMS and built-in OS search, and where a meaning-based, fully offline tool fits, including its limits.

Why small firms struggle to find documents across matters

Small firms struggle because their files are organised by matter, but lawyers remember documents by substance. You recall "the lease where the landlord carved out roof repairs", not Client-Smith/2023/Drafts/v4_final.docx.

A few patterns make this worse in small practices:

• Inconsistent naming. Different people, different years, different habits. Filenames rarely describe content.
• Scans and photos. Signed agreements, court stamps and client-supplied documents arrive as image-only PDFs with no searchable text.
• Saved emails. Correspondence exported as .msg or .eml files sits inside matter folders, invisible to anyone who doesn't know it's there.
• Knowledge leaves with people. When a paralegal moves on, so does their mental map of where things live.

Built-in tools help only partly. Windows Search, for example, relies on an index of file properties and contents that has to cover the right locations (Microsoft: search indexing in Windows), and like most OS search it matches words rather than ideas. If you don't recall the exact term the drafter used, you won't find the document.

What a local search tool for small law firms should do

A local search tool for small law firms should make every matter folder searchable by content while keeping all processing on the firm's own hardware. In practice that means a short checklist:

1. On-device processing. Text extraction, indexing and search should run locally. If the tool needs to upload documents to work, it isn't local, whatever the marketing says.
2. Works offline. The simplest test: disconnect from the internet and run a search. If it still works, your files aren't going anywhere.
3. Search by meaning, not just keywords. Semantic search matches the intent of a query rather than exact words, so "tenant responsible for structural repairs" can surface a clause phrased entirely differently. For more on the difference, see semantic vs keyword search.
4. The file types lawyers actually use. Word, PDF, Excel, saved emails, and scanned documents.
5. OCR for scans. Image-only PDFs need optical character recognition before they can be searched, and that should happen on the same machine.
6. Incremental updates. Matters change daily; re-indexing everything each time isn't practical.

The reason on-device matters so much for law is confidentiality. Attorney–client privilege is why lawyers guard client communications, and every additional third party that processes those files is another place to explain in an engagement letter or a client's security questionnaire.

Cloud DMS vs OS search vs a local semantic search tool

Each option solves a different problem: a cloud DMS manages documents, OS search finds filenames and words, and a local semantic tool finds content by meaning without moving files. Here is how they compare for a small practice:

These aren't mutually exclusive. A firm with a DMS may still want local search for archives that were never migrated. A firm without one may find that local search covers the real day-to-day pain, which is "where is that document?", without changing how files are stored. For a closer look at the legal use case, see how lawyers search case files by meaning.

How Fossick handles a firm's matter files on-device

Fossick, a desktop app for Windows and Mac that searches your local documents by meaning, entirely offline, is one example of this approach. You point it at folders, it builds an index on your machine, and you type what you remember about a document.

What happens under the hood:

• Everything runs locally. Text extraction, OCR, embedding and search all happen on the device. Documents are never uploaded, and it works with Wi-Fi off. Only licensing touches the internet.
• A small local model. It uses the open all-MiniLM-L6-v2 sentence-embedding model (384 dimensions, quantized, about 22 MB) to turn passages into vectors that can be compared by meaning.
• Legal-office file types. PDF, Word (.docx, .doc), Excel (.xlsx), PowerPoint (.pptx), OpenDocument text, plain text, Markdown, CSV, HTML, saved emails (.eml, .msg), and images (PNG, JPG).
• Scans become searchable. Scanned PDFs and images are run through on-device OCR automatically.
• Incremental re-indexing. Only new or changed files are re-processed.

For a sense of scale, in a measured test on an Apple M1 MacBook with 8 GB RAM, a folder of 52 mixed documents (Word, PDF including scanned pages, Excel, CSV, Markdown) was indexed from scratch in about 81 seconds, including model load and OCR. Searches over that folder took about 15-20 ms median, re-checking the unchanged folder took about 7 seconds, and the index was about 1.9 MB on disk. Larger matter archives will take longer to index the first time.

Just as important is what it is not: Fossick is search, not chat. There is no chatbot, no generative AI, and no summaries. It returns the documents and passages that match; reading them and forming the legal judgment stays with you.

Setting up cross-matter search in a small practice

The practical setup is to index the folders where matter files already live, rather than reorganising anything first. A workable approach:

1. Start with one high-value folder. Your precedents or templates folder, or a large active matter, shows quickly whether meaning-based search fits how you work.
2. Add closed-matter archives. This is where semantic search earns its keep, because nobody remembers what those files are called.
3. Include saved correspondence. If the firm exports key emails to matter folders as .msg or .eml, they become searchable alongside drafts and pleadings.
4. Let OCR process scans. Client-supplied scans and signed copies become text-searchable without a separate OCR step. See searching scanned PDFs locally for detail.
5. Write queries the way you'd describe the document to a colleague. "Indemnity capped at fees paid" or "letter refusing extension of time" usually beats a single keyword.

If matters sit on a shared drive, the same principle applies; searching a shared drive of contracts by meaning walks through it. Each person's index lives on their own computer, so keep existing access controls and device encryption in place.

Limits to understand before relying on it

A local search tool improves findability; it does not replace document management, security controls or professional judgment. Be clear-eyed about the gaps:

• Not a DMS. No version control, check-in/check-out, matter-centric metadata, or retention workflows.
• Per-device index. The index lives on the machine that built it. There is no central server enforcing who can search which matter, so rely on your existing folder permissions.
• Search depends on what's readable. Very poor scans or handwriting may OCR imperfectly, and password-protected files may not be indexed.
• No answers, only documents. Because it is search rather than generative AI, it won't draft, summarise or interpret. Many lawyers see that as a feature for privileged work, but it does mean you still read the results.
• Device security is still yours. On-device processing removes the cloud exposure, but laptops should still be encrypted and access-controlled.

For many small and solo practices, that trade-off is the right one: meaningful cross-matter search without a migration, a vendor holding client files, or a new system to administer.