@glubler/mcp-server
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., "@@glubler/mcp-server Search Glubler for a website called 'Hacker News' and glue it to my reading list."
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.
@glubler/mcp-server
Official Model Context Protocol (MCP) Server for Glubler.com.
Enables AI models (Claude Desktop, Cursor, GitHub Copilot, Zed, and custom LLM agents) to:
Search Glubler's catalog of registered websites, games, tools, and news.
Manage Glubler collections & subglublers (nested up to 7 levels deep).
Collect & Glue websites: When an AI suggests sites that don't exist in Glubler's database yet, Glubler automatically fetches, scrapes metadata, retrieves authentic PNG favicons, and auto-categorizes them.
Canva-Style Web Login: Log in or register through Glubler's web interface without copying tokens manually.
Available Tools
Tool Name | Description | Requires Auth |
| Check connection status & profile info | No |
| Start Canva-style browser login / registration flow | No |
| Clear saved credentials & disconnect | No |
| Search Glubler's | No |
| Fetch contents of any public Glubler list ( | No |
| List the authenticated user's Glubler tree & counts | Yes |
| View all glued websites and subglublers inside a specific list | Yes |
| Create a new Glubler collection or nested subglubler | Yes |
| Glue a single website to a list (auto-scrapes if missing) | Yes |
| Glue multiple websites at once (auto-scrapes & fetches missing) | Yes |
| Permanently delete a Glubler list | Yes |
| Remove a website from a Glubler list | Yes |
Related MCP server: Crawl4AI RAG MCP Server
Setup & Usage
1. Requirements
Node.js 20 or 22 LTS
npm, included with Node.js
A Glubler account
A Glubler MCP/API token
Verify that Node.js and npm are available:
node -v
npm -v2. Download and install
Download the repository ZIP from GitHub and extract it to a folder that is easy to remember, for example:
D:\glubler-mcp-serverOpen PowerShell and run:
cd "D:\glubler-mcp-server"
npm install
npm run buildnpm install installs the dependencies. npm run build compiles the TypeScript source into the dist folder.
The MCP server entry point will be:
D:\glubler-mcp-server\dist\index.jsYou can test the server directly with:
npm startThe server stays connected to its MCP client, so press Ctrl+C to stop a manual test.
3. Generate a Glubler API token
Sign in at glubler.com.
Open Settings.
Open MCP Server.
Select Generate MCP / API Token.
Copy the generated token.
Keep the token private. Never commit it to GitHub or paste it into a public issue or chat. The token expires after 30 days; generate a new one and restart the MCP connection when needed.
4. Configure OpenCowork
OpenCowork should start the MCP server as a local stdio process.
Open OpenCowork settings.
Go to the MCP server or integrations section.
Add a new local stdio MCP server.
Use these values, replacing the example path with your actual extraction path:
Setting | Value |
Name |
|
Type | STDIO |
Command |
|
Arguments |
|
Environment Variables | |
NOTION TOKEN | GLUBLER_AUTH_TOKEN |
ENTER VALUE | Your Glubler MCP/API token |
If OpenCowork asks for a single command line, use:
node "D:\glubler-mcp-server\dist\index.js"If OpenCowork asks for a JSON configuration, use:
{
"mcpServers": {
"glubler": {
"command": "node",
"args": ["D:\\glubler-mcp-server\\dist\\index.js"],
"env": {
"GLUBLER_API_URL": "https://glubler.com",
"GLUBLER_AUTH_TOKEN": "PASTE_YOUR_TOKEN_HERE"
}
}
}
}Save the configuration and restart the Glubler MCP connection. Then ask OpenCowork:
List my Glublers.
A successful response confirms that the server and token are working.
GLUBLER_AUTH_TOKEN is the recommended authentication method for OpenCowork. The glubler_login tool is only needed when no token is configured. If OpenCowork returns 401 Unauthorized, generate a fresh token in Settings → MCP Server, update GLUBLER_AUTH_TOKEN, and restart the connection.
5. Configure Claude Desktop
Add the following to your claude_desktop_config.json. Replace the path and token with your own values:
{
"mcpServers": {
"glubler": {
"command": "node",
"args": ["D:\\glubler-mcp-server\\dist\\index.js"],
"env": {
"GLUBLER_API_URL": "https://glubler.com",
"GLUBLER_AUTH_TOKEN": "PASTE_YOUR_TOKEN_HERE"
}
}
}
}Restart Claude Desktop after saving the configuration.
The browser login flow remains available as a fallback. It stores local credentials in ~/.glubler/auth.json, but a static token is recommended for unattended clients.
6. Configure other MCP clients
Use the same local stdio settings shown above. The important values are:
Command:
nodeArguments: the absolute path to
dist/index.jsEnvironment:
GLUBLER_API_URL=https://glubler.comEnvironment:
GLUBLER_AUTH_TOKEN=<your token>
For local backend development only, set GLUBLER_API_URL to http://localhost:5000 and use a token issued by that local backend.
Example Prompt for AI
Find 10 game news websites. Check Glubler first for ones already cataloged, choose popular gaming sites for the rest, create a Glubler collection named "Gaming Hub", and glue all 10 into it.
Glubler's backend will automatically fetch, scrape titles and favicons, auto-categorize new sites, and glue the accepted websites into your profile.
Troubleshooting
node or npm is not recognized
Install Node.js 20 or 22 LTS from nodejs.org, close and reopen PowerShell, then run:
node -v
npm -vnpm run build fails
Run the commands from the folder containing package.json:
cd "D:\glubler-mcp-server"
npm install
npm run buildDo not copy node_modules from another computer. Install dependencies on the machine that will run the server.
OpenCowork cannot find the server
Check that dist/index.js exists and that the path is absolute. Use double backslashes in JSON configuration.
Glubler returns 401 Unauthorized
The token is missing, expired, or incorrect. Generate a new token in Settings → MCP Server, update GLUBLER_AUTH_TOKEN, and restart the MCP connection.
Websites are rejected
Some sites are rejected because they are unreachable, time out, lack an authentic favicon, or do not meet Glubler's quality policy. Review the rejection reason returned by the batch tool.
Project Layout
src/ TypeScript source code
dist/ Generated JavaScript, created by npm run build
package.json Dependencies and npm scripts
package-lock.json Reproducible dependency versions
tsconfig.json TypeScript compiler settings
README.md Setup and usage documentationDo not commit or distribute node_modules, personal tokens, .env files, or local authentication files.
License
MIT License.
Available Tools
13 toolscreate_glublerA
Create a new Glubler (website list/collection) on the authenticated users profile.
Inputs: title (required), slugName, parentGlublerId.
delete_glublerA
Delete a Glubler collection and its child items permanently.
Inputs: glublerId (required).
get_glubler_detailsB
Get details of a specific Glubler including all glued websites and subglublers inside it.
Inputs: glublerId (required).
get_public_glublerA
Get a public Glublers contents by username and slug (no authentication required).
Inputs: slugName (required), username (required).
glubler_loginA
Start browser authentication only if GLUBLER_AUTH_TOKEN is not configured and no saved or active token is available. Token-based connections do not need this tool; use Glubler tools directly. If the API returns 401, ask the user to generate a new token in Glubler Settings > MCP Server, update GLUBLER_AUTH_TOKEN locally, and restart the MCP connection. Never ask the user to paste a token into chat.
Inputs: authUrl.
glubler_logoutA
Disconnect the authenticated Glubler account and remove saved local credentials.
Inputs: no named parameters.
glubler_whoamiA
Check connection status and return details of the currently authenticated Glubler user.
Inputs: no named parameters.
glue_websiteA
Glue/save a single website to a specific Glubler. If the website does not exist in Glubler yet, Glubler automatically scrapes metadata and fetches the authentic favicon.
Inputs: url (required), glublerId (required), title.
glue_websites_batchA
Glue/save multiple websites to a Glubler at once. For each URL, Glubler checks gDatabase or scrapes & fetches missing metadata/favicons automatically.
Inputs: urls (required), glublerId (required).
list_my_glublersA
Retrieve the authenticated users Glublers (bookmark collections/lists), including child subglublers and website count.
Inputs: no named parameters.
rename_glublerB
Rename a Glubler or subglubler (nested collection). Updates its display title and/or URL slug. Works for both top-level glublers and subglublers.
Inputs: glublerId (required), title, slugName.
search_glubler_catalogC
Search for existing websites and games/tools/news already cataloged in Glublers gDatabase.
Inputs: query (required), limit.
unglue_websiteB
Remove a website from a specific Glubler collection.
Inputs: url (required), glublerId (required).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 13 tools
Most tools map cleanly to distinct resources and actions, such as auth, listing, fetching details, and CRUD operations. The only real overlap is between glue_website and glue_websites_batch, and between listing vs. fetching details, but the descriptions are clear enough to guide correct selection.
The core CRUD and website tools mostly follow a readable verb_noun pattern, e.g., create_glubler, glue_website, unglue_website. The auth tools break the pattern with glubler_login, glubler_logout, and glubler_whoami, where the brand prefix comes first and whoami is not verb-first.
Thirteen tools is a well-scoped size for this domain. Each tool serves a clear purpose across authentication, discovery, collection lifecycle management, and website membership operations, with no obvious bloat.
The set covers the core bookmark/collection workflow well: auth, search, list, get, create, rename, delete, glue, batch glue, and unglue. Minor gaps exist, such as no direct move-website-between-collections operation or batch unglue, but these can be worked around with existing tools.
Related MCP Connectors
Web search, data extraction and controlled browser automation for AI assistants via OAuth.
Enables AI assistants to natively interact with the Serpzilla link-building marketplace.
Provides AI assistants with access to Seltz's powerful Web Search capabilities.
Web data for agents. Extract web pages, crawl websites, discover URLs, and search the web.
Related MCP Servers
- AlicenseBqualityFmaintenanceEnables AI assistants to manage StashDog inventory through natural language commands, supporting item management, collections, tags, smart search, and URL imports with secure authentication.11MIT
- FlicenseNot gradedqualityDmaintenanceEnables AI assistants to crawl websites, extract and store web content with semantic search capabilities using vector embeddings, and retrieve information through natural language queries with tag-based filtering and intelligent content cleaning.-
- AlicenseAqualityDmaintenanceEnables AI assistants to search, browse, and save to a semantic knowledge graph with vector search and hierarchical organization.51MIT
- FlicenseNot gradedqualityDmaintenanceEnables AI assistants to manage and search Obsidian notes, folders, metadata, and links directly.-