Skip to main content
Glama

Findling

Let your AI assistant search, read and cite your own files. Privately, on your own computer.

Tests License: AGPL-3.0 Python 3.12 macOS | Linux MCP server

Quick start · Connect your AI assistant · Documentation

Why Findling?

Findling lets your AI assistant work with your own files. Pick the folders you care about (documents, PDFs, web pages, notes), then just ask your assistant, for example Claude, a question. It searches those folders, reads the passages that matter and answers with quotes and links back to the source, so you can check every claim with one click. You don't need to know the exact words or where a file is: asking how much notice do I have to give on my flat? finds the right clause in your rental contract, even if it never uses the word "notice".

Why not just upload the files to the chat?

That works for a handful of documents, but not for the folders you actually keep:

Uploading files to a chat

Findling

How much

A chat only holds so much: a few documents

Thousands of files across all your folders

Finding the right file

You pick the files before you ask

Findling finds them for you

Answer quality

The assistant wades through everything you uploaded on every question and can miss the one detail that matters

The assistant only reads the passages that are relevant

Speed

The more you upload, the slower the chat and the sooner it hits its limits

A search takes about a second, so the assistant can look things up as often as it needs

Keeping up to date

Upload again for every new chat and after every change

Findling notices new and changed files by itself

Privacy

Whole files leave your computer

Files stay on your computer; the assistant only sees the passages it needs

Checking answers

Hard to tell where in a file a quote came from

Every quote links to the exact spot, highlighted

Findling also works on its own, as a search page in your browser: type what you're looking for, see the matching passages, and open each file right at that spot, highlighted.

Related MCP server: local-docs-mcp

Features

  • Finds by meaning and by exact words, in German, English and other languages, including IDs like INV-2024-0042

  • Reads PDFs, Word, PowerPoint, Excel, web pages, Markdown, notes and code

  • Watches your folders and only reprocesses what changed

  • Runs only on your computer and doesn't accept connections from other devices

Quick start

  1. Install uv, if you don't have it yet:

    curl -LsSf https://astral.sh/uv/install.sh | sh
  2. Install Findling:

    uv tool install --python 3.12 git+https://github.com/cedricfauth/findling
  3. Connect your AI assistant (see below).

  4. Add your folders: run findling (or let your assistant start it), open http://127.0.0.1:8765 and click Add folder…. Findling indexes them in the background.

  5. Ask your assistant about your files.

To update, run step 2 again with --force.

Connect your AI assistant

Findling is an MCP server, the open standard for connecting AI assistants to tools. Your assistant gets four tools: search, read (read on from a passage or open a section such as § 1599), browse (folders, files, sections) and open (show a passage in your browser).

Claude Desktop (macOS): download findling.mcpb from the latest release and double-click it. Claude then starts Findling in the background whenever it needs it.

Claude Code:

claude mcp add findling -- findling-mcp

Other apps that support MCP should work too. Most take a config like this (use the full path that which findling-mcp shows; some apps name the outer key differently, see their MCP docs):

{
  "mcpServers": {
    "findling": { "command": "/Users/you/.local/bin/findling-mcp" }
  }
}

Like with Claude, the command starts Findling in the background when needed. Apps that connect over HTTP instead can use http://127.0.0.1:8765/mcp while Findling runs. If an app doesn't work with it, please open an issue.

Web chats such as claude.ai or ChatGPT in the browser can't reach Findling: it only listens on your own computer. More details: Using Findling.

examples/findling-recht is an example Claude skill for German law that makes Claude ground its answers in indexed laws and court decisions and cite every source. To make one for your own field, use examples/findling-skill-creator: it studies your indexed folders and writes the skill.

Use it on its own

findling                                      # web app at http://127.0.0.1:8765
findling-search "how do I restore a backup"   # search from the terminal

Platforms

Works

GPU for indexing

Claude Desktop

macOS, Apple Silicon

Yes

Automatic (untested)

Yes

macOS, Intel

Yes

No, CPU only

Yes

Linux

Yes

NVIDIA with an extra install (untested)

Not available

Windows

Not yet

Status

Version 0.1.0, the first public release. Known limits:

  • GPU indexing hasn't been tested on real hardware yet. If it causes trouble, set FINDLING_DEVICE=cpu and open an issue.

  • Images yield no text. Scanned PDFs need OCR switched on (off by default).

  • No Windows support yet.

Ideas for later are in docs/IMPROVEMENTS.md.

Documentation

  • Using Findling: the web app, command-line search, and the MCP tools in detail

  • Configuration: settings, folders, Linux, GPU, HTML pages and extensions

  • How it works: parsing, embeddings, storage and how results are ranked

  • Development: working on Findling, tests and releases

License

Copyright (C) 2026 Cedric Fauth.

Findling is free software under the GNU Affero General Public License v3.0 or later. In short: you may use, change and share it, but whoever shares it, or lets others use a changed version over a network, must make the source available under the same license. For use in a closed product, a commercial license is available on request.

The bundled pdf.js viewer is licensed under the Apache License 2.0.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      D
      maintenance
      Enables AI assistants to perform semantic, hybrid, and filtered search on indexed local documentation with RAG capabilities.
      2
      MIT
    • F
      license
      A
      quality
      D
      maintenance
      Enables AI assistants to search, read, and retrieve context from local knowledge bases with full-text search, absolute paths, and section-level details.
      4
      -
    • A
      license
      Not graded
      quality
      C
      maintenance
      Enables private, offline semantic search across local files (documents, images, videos) using OCR and vector search, and optionally performs web research with cited sources.
      AGPL 3.0