Skip to content
Avlok Docs

Overview

Avlok exposes your photo library to AI assistants over MCP, the open Model Context Protocol. This page explains what that means and how the pieces fit.

What MCP gives you

Section titled “What MCP gives you”

MCP is the standard that assistants like Claude and ChatGPT use to reach external tools. When you connect Avlok, your assistant gains a set of tools it can call on your behalf: search your photos by description, look up a photo’s details, find pictures from a place or trip, and more.

You stay in the conversation. Instead of switching to the app and scrolling, you ask, and the assistant pulls from your library and shows you what it found.

Two ways to connect

Section titled “Two ways to connect”

Avlok offers two connection paths. They differ in where the server runs and, as a result, in how much touches the cloud.

Local, for Claude Desktop

Avlok runs a small server on your own computer. Your assistant reaches your library directly, with nothing sent to our servers to make it work. This is the most private path.

Cloud, for web assistants

Turn on Cloud AI Access and Avlok serves your library from the cloud, so Claude on the web, ChatGPT, and other assistants can reach it from any device. You choose which folders to include.

How the local path works

Section titled “How the local path works”

The local server is a small program bundled with the Avlok desktop app. You enable it from Settings → AI Integration, which gives you a toggle and a button to copy the configuration your assistant needs.

It reads your local search index and cached thumbnails. It is read-only: it can find and describe photos, never change or delete them. It makes no outbound network calls of its own. Everything happens on your machine.

How the cloud path works

Section titled “How the cloud path works”

When you turn on Cloud AI Access, the folders you choose are made available to a hosted Avlok server. Your assistant connects to it as a custom connector, authenticated with your Avlok account over OAuth.

The hosted server holds small thumbnails and short text descriptions of the photos in your chosen folders, enough for an assistant to search and show results, never your full-resolution images. See How Avlok handles your data for the specifics.

What your assistant can do

Section titled “What your assistant can do”

Once connected, the assistant has tools to:

  • Search your photos by plain-language description
  • Filter by date, place, camera, and other details your photos carry
  • Look up the full details of a single photo
  • Find photos taken near a location
  • Browse photos from a detected trip
  • Reference people you have named, and the places and things in your library
  • Find visually similar photos
  • Search the words spoken in your videos (your computer does the listening; with Cloud AI Access the words are searchable from every device and connected assistant)

Each of these maps to a specific tool, documented in the Tool reference.

Next steps

Section titled “Next steps”