MCP (Model Context Protocol)

Bookmarkjar MCP connects your bookmark library to an AI assistant. After a one-time browser sign-in, you can ask the assistant to find saved links, add new bookmarks, or prepare a cleanup review.

MCP access requires an active paid plan. You do not need to create or paste an API key.

Choose your AI app

The Bookmarkjar server URL is:

https://bookmarkjar.com/api/mcp

Claude Code

Run this command in a terminal:

claude mcp add --transport http bookmarkjar https://bookmarkjar.com/api/mcp

Start a new Claude Code session, then ask:

Search my Bookmarkjar bookmarks for design systems.

In Claude Code, type /mcp, select bookmarkjar, and choose Authenticate. Claude opens Bookmarkjar in your browser. Sign in, review the permissions, click Allow, and return to Claude Code.

Claude on the web or Claude Desktop

Open Claude → Settings → Connectors.
Click Add custom connector.
Name it Bookmarkjar and paste https://bookmarkjar.com/api/mcp.
Under Authentication, keep Always required.
Under OAuth client, choose No client ID — register one automatically. Bookmarkjar supports Dynamic Client Registration (DCR); do not choose Anthropic-hosted client metadata.
Save the connector, click Connect, and complete the Bookmarkjar browser authorization.
Enable Bookmarkjar in a conversation from the + menu under Connectors.

Team and Enterprise workspaces require an owner to add the custom connector before members can connect it.

ChatGPT

In ChatGPT, open Settings → Security and login and enable Developer mode.
Open ChatGPT Plugins and click the + button.
Enter a name and description, then choose a public MCP connection.
Use https://bookmarkjar.com/api/mcp as the server URL.
Keep authentication required. If ChatGPT asks how to register the OAuth client, choose automatic registration or DCR; do not enter an API key or client secret.
Create the connection and review the discovered tools.
Add Bookmarkjar from the tools menu in a new conversation, then complete the Bookmarkjar browser authorization.

Developer mode and custom MCP connections depend on your ChatGPT plan and workspace policy.

Cursor

Create .cursor/mcp.json in your project, or ~/.cursor/mcp.json to use Bookmarkjar in every project:

{
  "mcpServers": {
    "bookmarkjar": {
      "url": "https://bookmarkjar.com/api/mcp"
    }
  }
}

Reload Cursor, open Customize → MCP, enable bookmarkjar, and choose Authenticate. Cursor discovers Bookmarkjar OAuth automatically through DCR; no static client ID or API key is required.

Codex CLI and ChatGPT desktop

Add this to ~/.codex/config.toml on macOS or Linux, or %USERPROFILE%\.codex\config.toml on Windows. ChatGPT desktop, Codex CLI, and the Codex IDE extension share this configuration:

[mcp_servers.bookmarkjar]
url = "https://bookmarkjar.com/api/mcp"
auth = "oauth"

Run these commands:

codex mcp login bookmarkjar
codex mcp list

Complete the Bookmarkjar browser authorization, then restart Codex or ChatGPT desktop.

Let your AI set it up

Paste the prompt below into Claude, ChatGPT, Cursor, Codex, or another capable assistant. It will detect the client, guide you through the correct setup, and stop for the browser authorization step.

Help me connect this AI client to Bookmarkjar MCP.

Bookmarkjar MCP server:
https://bookmarkjar.com/api/mcp

Rules:
- Detect whether I am using Claude Code, Claude on the web/Desktop, ChatGPT, Cursor, Codex CLI, or another MCP client.
- Tell me what you detected before changing anything.
- Do not ask me for a Bookmarkjar API key. This connection uses browser-based OAuth.
- Ask before editing a configuration file.

Use the matching setup:

Claude Code:
Run:
claude mcp add --transport http bookmarkjar https://bookmarkjar.com/api/mcp
Then tell me to type /mcp, select bookmarkjar, choose Authenticate, and complete browser authorization.

Claude on the web or Claude Desktop:
Guide me to Settings → Connectors → Add custom connector. Use the server URL above. Keep Authentication set to Always required. Under OAuth client, choose No client ID — register one automatically (DCR), not Anthropic-hosted client metadata. Tell me that Team or Enterprise workspaces require an owner to add the connector first.

ChatGPT web:
Guide me to Settings → Security and login → Developer mode. Then open ChatGPT Plugins, click +, enter a name and description, choose a public MCP connection, and use the server URL above. Keep authentication required. If asked for OAuth client registration, choose automatic registration or DCR. Mention that availability depends on my plan and workspace policy.

Cursor:
Ask before creating or updating .cursor/mcp.json for the project, or ~/.cursor/mcp.json globally. Add:
{
  "mcpServers": {
    "bookmarkjar": {
      "url": "https://bookmarkjar.com/api/mcp"
    }
  }
}
Then tell me to reload Cursor, open Customize → MCP, enable bookmarkjar, and choose Authenticate.

Codex CLI or ChatGPT desktop:
Ask before updating my Codex config. Add this to ~/.codex/config.toml on macOS/Linux or %USERPROFILE%\.codex\config.toml on Windows:
[mcp_servers.bookmarkjar]
url = "https://bookmarkjar.com/api/mcp"
auth = "oauth"
Then run codex mcp login bookmarkjar followed by codex mcp list.

After registration:
1. Tell me to start a fresh chat or session.
2. Ask the client to search my Bookmarkjar bookmarks for "design systems".
3. Wait while I sign in to Bookmarkjar and click Allow in the browser.
4. Verify that the bookmarks.search tool returns a result or an empty result without an authorization error.
5. Report what was configured and ask what I want to search or save next.

Things to try

Once connected, speak normally:

Search my Bookmarkjar bookmarks for PostgreSQL performance.

Save https://example.com to my Bookmarkjar bookmarks.

Create a cleanup review for duplicate design bookmarks.

A cleanup request only creates a review. Bookmarkjar does not delete candidates until you approve them in the Cleanup screen.

What access you approve

The browser consent screen asks for one or both permissions:

Read: search your bookmarks.
Write: save bookmarks and create cleanup reviews.

Access is tied to your signed-in Bookmarkjar account. You can deny the request or disconnect the integration from your AI client.

Available tools

bookmarks.search — search and filter your bookmarks.
bookmarks.insert — save one or more URLs.
cleanup.create — prepare a cleanup review without automatically deleting anything.

The server uses Streamable HTTP and browser-based OAuth at https://bookmarkjar.com/api/mcp. Do not add an API key header to this endpoint.

Troubleshooting

No browser window opens: start a fresh chat or client session, confirm bookmarkjar is enabled, then call a Bookmarkjar tool again.
401 Unauthorized: reconnect Bookmarkjar and finish the browser authorization.
403 insufficient_scope: reconnect and approve the requested read or write permission.
Subscription required: connect with a Bookmarkjar account that has an active paid plan.
ChatGPT cannot add the server: confirm Developer mode is available for your plan and allowed by your workspace.
Claude Team or Enterprise cannot add the connector: ask a workspace owner to add it under organization connector settings.
Redirect or signature error: remove the connection, add it again, and complete a fresh authorization request.

Getting started

Bookmarks

Advanced Topics

More

Choose your AI app

Claude Code

Claude on the web or Claude Desktop

ChatGPT

Cursor

Codex CLI and ChatGPT desktop

Let your AI set it up

Things to try

What access you approve

Available tools

Troubleshooting

​Choose your AI app

​Claude Code

​Claude on the web or Claude Desktop

​ChatGPT

​Cursor

​Codex CLI and ChatGPT desktop

​Let your AI set it up

​Things to try

​What access you approve

​Available tools

​Troubleshooting

Choose your AI app

Claude Code

Claude on the web or Claude Desktop

ChatGPT

Cursor

Codex CLI and ChatGPT desktop

Let your AI set it up

Things to try

What access you approve

Available tools

Troubleshooting