linkwarden-mcp
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@linkwarden-mcp list my collections"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
linkwarden-mcp
An MCP server that wraps the
Linkwarden bookmark API in a small set of token-frugal
tools. Tool responses are compact plain text (#id name :: url) so listing
collections or links costs as few tokens as possible.
Tools
Tool | What it does |
| List collections as |
| List links as |
| Check if a URL already exists (tracking params stripped first). |
| Add a link: strips tracking params, skips duplicates, then moves it into a collection. |
| Move an existing link to another collection. |
| Delete a link by id. |
| Create a new collection. |
Behaviour baked in from experience
Tracking params are stripped on add (
utm_*,fbclid,gclid,gbraid,gad_source,gad_campaignid,rcm, …); meaningful params likevariantandmodelare kept.Two-step create. Linkwarden ignores
collectionIdon link creation, soadd_linkPOSTs the link then PUTs it into the target collection.Correct PUT shape. Updates always send
id,url,collection.{id,ownerId}andtags, which the API requires.Full-list dedup. Linkwarden caps page size at ~50, so dedup walks every page via cursor pagination rather than trusting a single
limit.
Related MCP server: rainkeeper
Configuration
Configuration comes from environment variables — no secret is ever hardcoded:
Variable | Required | Default | Notes |
| no |
| Base URL, no |
| yes | — | API token (Settings → Access Tokens). Secret. |
| no |
| Owner id used when moving links. |
How the token is stored: in production it is supplied by the MCP host via the
server config's env block (see below) — it is not read from disk. For local
development you may instead copy .env.example to .env and fill in the token;
.env is git-ignored and loaded automatically. Host-provided env vars take
precedence over .env.
Run
uv sync
uv run linkwarden-mcp # serves over stdioRegister with an MCP host
Claude Code:
claude mcp add linkwarden -e LINKWARDEN_URL=http://links.lan -e LINKWARDEN_TOKEN=your-token -- uv --directory C:/Users/olive/Github/linkwarden-mcp run linkwarden-mcpOr Claude Desktop (claude_desktop_config.json):
{
"mcpServers": {
"linkwarden": {
"command": "uv",
"args": ["--directory", "C:/Users/olive/Github/linkwarden-mcp", "run", "linkwarden-mcp"],
"env": {
"LINKWARDEN_URL": "http://links.lan",
"LINKWARDEN_TOKEN": "your-token"
}
}
}
}Development
uv run ruff check . # lint
uv run ruff format . # format
uv run ty check # type checkAvailable Tools
7 toolsadd_linkA
Add a link. Tracking params are stripped; by default duplicates are skipped.
Creates the link then moves it into `collection_id` (the API ignores the collection on
creation, so a follow-up update is required). Pass collection_id=None to leave it in Unorganized.
Inputs: url (required), collection_id, name, tags, +1 more. Returns: result
create_collectionB
Create a new collection and return its #id name.
Inputs: name (required), is_public. Returns: result
delete_linkB
Permanently delete a link by id.
Inputs: link_id (required). Returns: result
find_linkA
Check whether a URL already exists (tracking params stripped). Returns matches or (none).
Inputs: url (required). Returns: result
list_collectionsA
List all collections as #id name (link count), one per line.
Inputs: no named parameters. Returns: result
list_linksB
List links as #id name :: url. Omit collection_id for the newest links across all collections.
Inputs: collection_id, limit. Returns: result
move_linkC
Move an existing link into a different collection.
Inputs: link_id (required), collection_id (required). Returns: result
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 7 tools
Each tool has a clear, distinct purpose: add, delete, find, move links; create and list collections; list links. There is no overlap or ambiguity.
All tool names follow a consistent verb_noun pattern in snake_case (e.g., add_link, list_collections), making it easy for an agent to infer functionality.
With 7 tools, the set is well-scoped. It covers core operations for a link management service without being too sparse or bloated.
The set covers primary CRUD operations for links (add, delete, move, find) and collection management (create, list). Missing update operations for links/collections and deletion of collections are minor gaps.
Related MCP Connectors
Search, save, tag, and summarize the links in your Link Crypt bookmark library.
Save, search and organize bookmarks, highlights, feeds and knowledge cards in a Linkflare library.
All your bookmarks in one library. Share them with your AI agents. Read-only search over your saves.
Search and chat with your X bookmarks, with citations back to the original posts.
Related MCP Servers
- FlicenseAqualityDmaintenanceEnables to manage Raindrop.io bookmarks programmatically—add, search, and organize bookmarks from LLM apps.415-
- AlicenseBqualityCmaintenanceExposes the Raindrop.io API as Claude tools for managing bookmarks, collections, tags, and bulk operations, with proper collection move support.191MIT
- AlicenseAqualityBmaintenanceEnables MCP clients to search, organize, and read preserved article text from a Linkwarden bookmark manager, supporting both read and write operations with safety confirmations.2880 npm1MIT
- AlicenseAqualityBmaintenanceEnables searching and managing Raindrop.io bookmarks via the official REST API, including listing collections, retrieving bookmarks, and creating or updating bookmarks.6MIT