A tiny, stdlib‑only bookmark manager inspired by the Unix philosophy and pass:
- One text file per bookmark (
.bm) with front matter + freeform notes - Human‑readable paths with a short hash to avoid collisions
- Stable IDs derived from the URL (rename‑safe)
- Greppable store; composable CLI
- Atomic writes & path‑safety checks
- JSON / JSONL output for pipelines
- Netscape HTML import/export for browser interoperability
- Optional Git sync for history and cross‑device
Works on macOS, Linux, WSL, and Windows (PowerShell). No third‑party dependencies.
Most bookmark tools are databases or browser‑locked. bm chooses text first: plain UTF‑8 files that last decades, are easy to diff, and play well with your editor, shell, and Git. It embraces "do one thing well" and stays small so you can integrate it anywhere.
Requires Python >=3.8. Install from PyPI:
pip install bkmrkAlternatively, using uv (a fast Python package installer):
uv pip install bkmrkFor the latest development version, clone and install locally:
git clone https://github.com/jtabke/bkmrk
cd bkmrk
python -m pip install .Or with uv:
git clone https://github.com/jtabke/bkmrk
cd bkmrk
uv pip install .Development workflows can pull in the optional extras declared in pyproject.toml:
python -m pip install -e '.[dev]'Or with uv:
uv pip install -e '.[dev]'Prefer running straight from the repository? The module entry point works without installation:
python -m bm --helpOn Windows (PowerShell):
python -m bm --help# initialize a new store (optionally a git repo)
bm init --git
# import bookmarks from a browser export (Netscape HTML)
bm import netscape ~/Downloads/bookmarks.html
# add a bookmark
bm add https://example.com -n "Example" -t ref,demo -d "Short note"
# list newest bookmarks (ID, path, title, URL)
bm list
# search across title/url/tags/body
bm search kernel
# search within a specific path
bm search kernel --path dev/linux
# list directory prefixes
bm dirs
# open the first result
ID=$(bm search kernel --jsonl | head -1 | jq -r '.id')
bm open "$ID"
# export for browsers (Netscape HTML)
bm export netscape > bookmarks.htmlDefault store directory is ~/.bookmarks.d (override via $BOOKMARKS_DIR). Each bookmark is a single .bm file; directories serve as namespaces.
~/.bookmarks.d/
dev/python/fastapi-3a1b2c4.bm
news-ycombinator-com-1234567.bm
README.txt
File names are human readable and end with a short hash of the URL to avoid collisions.
Each .bm file contains front matter and an optional body:
---
url: https://example.com/blog/post
title: Great post
tags: [read, blog, "needs,comma"]
created: 2025-09-16T08:42:00-07:00
modified: 2025-09-17T09:10:00-07:00
---
Longer notes, checklists, code blocks…
tagsis a list and supports quoting for commas/spaces- Any extra keys are preserved on round‑trip
Each bookmark has a stable ID derived from its URL (BLAKE2b short hash). The ID does not change if you rename/move the file. You can use either the ID or a path‑like slug with commands.
Run bm --help or bm <command> --help for command details.
Create a store; optional --git initializes a Git repo.
bm init --gitAdd a bookmark. --edit opens your $EDITOR with a pre‑filled template.
bm add <url> [-n TITLE] [-t tag1,tag2] [-d NOTES] [-p dir1/dir2] [--id SLUG] [--edit] [-f]Prints the stable ID on success.
List bookmarks (newest first).
bm list [--host HOST] [--since ISO|YYYY-MM-DD] [-t TAG] [--path PREFIX] [--json|--jsonl]Full‑text search across title, url, tags, and body.
bm search <query> [--path PREFIX] [--json|--jsonl]Display metadata/notes or open the URL in your default browser:
bm show <ID|path>
bm open <ID|path>bm edit <ID|path> # bumps modified timestamp
bm rm <ID|path>
bm mv <SRC> <DST> [-f]List discovered tags (from folder segments and header tags), or mutate tags without opening an editor.
bm tags
bm tag add <ID|path> tag1 tag2
bm tag rm <ID|path> tag1List all known directory prefixes in the bookmark store.
bm dirs [--json]Merge duplicate bookmarks that resolve to the same normalized URL. The command unions tags (including folder segments), keeps the most informative entry, and appends any extra notes with provenance markers.
bm dedupe [--dry-run] [--json]--dry-runprints the planned merges without modifying files--jsonemits a machine-readable summary of the merge actions
Netscape HTML (for browsers) and JSON exports; Netscape import with folder hierarchies preserved.
bm export netscape [--host HOST] [--since ISO|YYYY-MM-DD] > bookmarks.html
bm export json > dump.json
bm import netscape bookmarks.html [-f]If the store is a Git repo, stage/commit and (if upstream exists) push.
bm sync--hostmatches the URL host (case‑insensitive, ignores leadingwww.)--pathfilters by path prefix (e.g.,--path dev/pythonshows only entries under that directory tree)--sinceacceptsYYYY-MM-DDor full ISO timestamps; comparisons are proper datetimes--jsonemits a single JSON array;--jsonloutputs one JSON object per line (NDJSON)
Common JSON schema fields: id, path, title, url, tags, created, modified.
fzf launcher
bm list --jsonl | fzf --with-nth=2.. | awk '{print $1}' | xargs -r bm openOpen the latest saved from a host
bm list --host example.com --jsonl | head -1 | jq -r '.id' | xargs -r bm openRofi launcher
#!/bin/sh
choice="$(
bm list --jsonl |
jq -r '.id + "\t" + .title + " — " + .url' |
rofi -dmenu -i -p "bm"
)"
if [ -n "$choice" ]; then
bm open "$(printf "%s" "$choice" | cut -f1)"
fiSave as bm-rofi.sh, make it executable (chmod +x bm-rofi.sh), and bind it to a hotkey. The
tab delimiter keeps IDs intact even when titles contain spaces; rofi shows the full title and URL
while bm open receives only the bookmark ID.
dmenu launcher
#!/bin/sh
choice="$(
bm list --jsonl |
jq -r '.id + "\t" + .title + " — " + .url' |
dmenu -l 15 -i -p "bm"
)"
if [ -n "$choice" ]; then
bm open "$(printf "%s" "$choice" | cut -f1)"
fiYou can adjust -l 15 to change the number of visible rows. Because the script uses tabs between
the ID and the description, cut -f1 reliably extracts the ID even when titles or URLs contain
spaces.
Bulk tag HN links
bm list --host news.ycombinator.com --jsonl | jq -r '.id' | xargs -n1 bm tag add hnList bookmarks in a specific category
bm list --path dev/python
bm search "framework" --path devExplore directory structure
bm dirs
bm dirs --json | jqExport → browser import
bm export netscape > ~/Desktop/bookmarks.html
# Import that file in your browser’s bookmarks managerSync with Syncthing
For cross-device synchronization without Git, use Syncthing to sync your bookmark store:
- Install Syncthing on all devices.
- Add your bookmark store directory (
~/.bookmarks.dor$BOOKMARKS_DIR) as a synced folder in Syncthing. - Configure devices to share the folder bidirectionally.
- Syncthing will keep your bookmarks in sync across devices automatically.
Auto-export for browser import
To automatically export bookmarks to Netscape HTML for browser import:
#!/bin/bash
# auto_export.sh
bm export netscape > ~/bookmarks_auto.html
echo "Bookmarks exported to ~/bookmarks_auto.html. Import this file in your browser."Run this script periodically or on demand to generate an up-to-date bookmark file for browser import.
- Store directory: set
BOOKMARKS_DIRor pass--storeto any command - Editor:
VISUALorEDITOR(supports commands likecode --wait)
Windows notes:
- Paths avoid reserved names and use atomic replaces; long paths depend on OS settings
- Atomic writes: all modifications write to a temp file then
os.replaceit - Path safety:
..and absolute paths are rejected; files cannot escape the store - No network by default:
bmnever fetches content (future hooks can) - Git: pushes only if an upstream is configured
# lint (optional) — stdlib only, so just run the script
python3 -m compileall src
# run tests (if added)
pytest -qbm reindex+ optional on‑disk index for very large stores- Markdown/CSV exports
- Simple HTTP UI (
bm serve) and browser extension hooks - Optional encryption (GPG or git‑crypt) for private notes
MIT. Do what you want; a credit is appreciated.