MemoReef doesn't try to replace your vault — it feeds it. Browser pages, highlights, bookmarks, documents, chat links, and video archives all become Markdown Drops with frontmatter your vault, Obsidian's graph view, and Dataview already understand. There's no separate app to keep in sync and nothing proprietary to export later.
Import your bookmarks
Point the CLI at any existing Obsidian vault and MemoReef writes directly into it.
$ python3 -m memoreef.cli import bookmarks.html --vault /path/to/your-vault
$ python3 -m memoreef.cli import-links links.txt --vault /path/to/your-vault
$ python3 -m memoreef.cli import-csv links.csv --vault /path/to/your-vault
bookmarks.html is a standard Netscape-style export from Chrome, Brave, Arc, Firefox, or Safari. links.txt is one URL per line. links.csv accepts title,url,source,tags columns. Replace /path/to/your-vault with your real vault folder.
Capture from your browser
The Drop to Reef browser extension sends the current page — plus any highlighted text — to your local MemoReef server as a new Drop. Install it via developer mode from the repository under extensions/drop-to-reef/. Configure your server URL and Capture Token once; the extension remembers both.
$ memoreef serve --vault /path/to/your-vault
The server accepts writes only with a valid Capture Token stored at ~/.config/memoreef/capture-token. Stop it when you are not actively capturing. A localhost bookmarklet is also available for simpler setups — see the README.
Chat commands and video archives
Gateway adapters for Telegram, Discord, WhatsApp, and iMessage-style bots can forward commands like /reef, /drop, youtube:, or signal: to MemoReef's /api/capture endpoint. Each message becomes a Drop in your vault.
Tokwise short-form video archives import with transcript, summary, metadata, and classification attached:
$ memoreef import-tokwise ~/.tokwise/videos/videos.jsonl --vault /path/to/your-vault
All sources — regardless of how they arrive — land in MemoReef/Drops/ with frontmatter recording their import source and status.
What lands in your vault
One Markdown file per bookmark, written into a predictable folder:
your-vault/
└── MemoReef/
└── Drops/
├── local-ai-agents-for-small-teams.md
├── another-saved-source.md
└── ...
Folder paths from your browser's bookmark structure are preserved as frontmatter, not as nested folders — so importing a few hundred bookmarks doesn't leave you with a hundred-level-deep tree.
Example frontmatter
Every Drop looks like this — readable on its own, queryable by Dataview, and parseable by an agent:
---
title: "Local AI Agents for Small Teams"
url: "https://example.com/local-agents"
type: drop
status: drift
import_source: browser-extension
agent_ready: true
treasure: false
folders:
- "AI Agents"
tags:
- ai-agents
---
# Local AI Agents for Small Teams
Source: [https://example.com/local-agents](https://example.com/local-agents)
## Summary
_Not enriched yet._
## Notes
## Agent Brief
- Status: Drift
- Suggested next action: triage this Drop.
How to use MemoReef as a local agent memory layer for Obsidian
MemoReef turns your saved bookmarks into structured Markdown files that local agents can search, cite, and build on inside your Obsidian vault.
Agent-ready Obsidian source library
status: drift and agent_ready: true aren't decoration — they're queryable signal. A Dataview query over MemoReef/Drops can list everything still unsorted, everything marked as Treasure, or everything tagged to a project. A local agent reads the same fields to decide what to triage next, without you writing a single line of glue code.
If you use the Dataview plugin, imported Drops are queryable immediately:
```dataview
TABLE url, folders
FROM "MemoReef/Drops"
WHERE status = "drift" AND agent_ready = true
```
This lists unsorted Drops that are ready for review or agent handoff.
Honest status
MemoReef is a command-line tool today, not a published Obsidian community plugin. It writes Markdown files into your vault folder; it doesn't run inside Obsidian's UI, live-sync, or show up in Community Plugins. A more native plugin and deeper Dataview conventions are on the roadmap — not built yet, so there's nothing to install from inside Obsidian.
Try it yourself
MemoReef is open source and runs entirely on your machine.