# Overview

What Bookmark Context is and how the pieces fit together.

Bookmark Context saves web pages into named collections, indexes their text
content, and makes those collections searchable from an AI client over MCP.

## The problem it solves

An AI assistant can't see the pages you've bookmarked. Bookmark Context closes
that gap: you save a page from the browser, the daemon extracts and embeds its
content, and your assistant retrieves the relevant passages on demand.

## The pieces

| Component | Command | Role |
|---|---|---|
| Daemon | `bookmark-context serve` | REST API on `127.0.0.1:7331`. Owns the database and the indexing pipeline. |
| Chrome extension | — | Side-panel UI. Creates collections, saves the current tab, shows index status. |
| MCP server | `bookmark-context mcp` | Spawned by your editor. Reads the same stores directly and answers tool calls. |

The daemon and the MCP server share two on-disk stores — a SQLite database and a
Chroma vector store — but never talk to each other. The extension reaches the
daemon over authenticated HTTP; the MCP server bypasses HTTP entirely.

## The 30-second picture

1. Create a collection in the side panel.
2. Open a page, click **Save to collection**. The daemon fetches the page, scans
   it for prompt-injection, splits it into chunks, embeds them, and stores the
   vectors.
3. In your AI client, ask a question. It calls `list_collections` to find the
   collection, then `search_collection` or `ask_collection` to pull the passages
   that answer you.

## Next

- [Requirements](/docs/getting-started/requirements/)
- [Install](/docs/getting-started/install/)
- [First run](/docs/getting-started/first-run/)