Integration

Connect ContextBolt to Claude Code via MCP

Beginner 2 minutes setup By David Hamilton

Claude Code is Anthropic’s command-line interface for Claude. It runs in your terminal, reads your codebase, and can use MCP servers to connect to external tools and data — including your ContextBolt bookmarks.

This integration is particularly useful for developers who save technical content across Twitter/X, Reddit, and LinkedIn. Instead of context-switching to your browser to look up that article about caching strategies or that Reddit thread about Kubernetes networking, you can ask Claude Code to search your bookmarks directly from the terminal.

Why this matters for coding workflows

Developers hoard bookmarks. Tweets about architectural patterns, Reddit discussions about library trade-offs, LinkedIn posts about scaling lessons. The problem is finding them when you actually need them.

With the Claude Code integration, your saved content becomes part of your coding context. You’re debugging a tricky issue, and you remember saving something relevant last month — just ask Claude Code to find it. The search is semantic, so you describe what you’re looking for in plain language rather than trying to remember exact keywords.

This is especially powerful because Claude Code already understands your codebase. It can combine what it knows about your code with what you’ve saved, making your bookmarks genuinely useful during development rather than collecting dust in a browser sidebar.

Configuration file alternative

If you prefer editing config files directly, add this to ~/.claude/settings.json:

{
  "mcpServers": {
    "contextbolt": {
      "command": "npx",
      "args": ["-y", "@contextbolt/mcp-server"],
      "env": {
        "CONTEXTBOLT_TOKEN": "your-token-here"
      }
    }
  }
}

Or for project-level config, create .mcp.json in your project root with the same structure. Project-level config is useful if your team shares a ContextBolt setup.

Tips for developers

Search before you search the web. Before reaching for Google or Stack Overflow, check your bookmarks. You’ve already curated high-quality content — use it first.

Combine with code context. Ask Claude Code something like “I’m refactoring the auth module. Do I have any saved content about OAuth best practices?” Claude can connect your bookmarks to the code it’s looking at.

Use platform filters for technical content. Reddit saves tend to be deep technical discussions. Twitter saves tend to be quick tips and opinions. Filtering by platform helps narrow results when you know where you saved something.

Setup guide

  1. Get your MCP token from ContextBolt

    Open the ContextBolt extension popup in Chrome and click Settings. Under the MCP section, click 'Copy MCP Token'. You need a Pro subscription (£4/month) as MCP is a Pro feature.

  2. Add ContextBolt via the CLI

    Open your terminal and run: claude mcp add -e CONTEXTBOLT_TOKEN=your-token-here contextbolt -- npx -y @contextbolt/mcp-server. This registers the server at the user level so it's available in all your projects. Replace your-token-here with the token you copied.

  3. Test the connection

    Start a new Claude Code session and ask 'Search my bookmarks for articles about API design'. If the connection is working, Claude will call the ContextBolt MCP tools and return matching bookmarks from your collection.

Example prompts

Once connected, try asking Claude Code:

Search my bookmarks for anything about database migrations

While working on a migration, pull in articles and threads you previously saved about migration strategies, rollback patterns, or schema versioning.

What did I save about React server components?

When building a feature, surface your saved reference material without leaving the terminal. Claude searches semantically, so it finds relevant content even with different wording.

Find my Reddit saves about Rust error handling

Filter by platform and topic. Useful when you saved a detailed Reddit discussion months ago and want to reference it mid-session.

I'm debugging a rate limiter. Do I have any saved content about rate limiting patterns?

Use this pattern when you're stuck. Claude searches your bookmarks for anything related to your current problem and brings relevant saves into the conversation.

What you can do

Search bookmarks by meaning during coding sessions

Filter by platform (Twitter, Reddit, LinkedIn)

Pull saved reference material into terminal conversations

Access bookmarks from any project directory

Combine bookmark search with code analysis in the same session

Works alongside other MCP servers you have configured

Frequently asked questions

What's the difference between project-level and user-level MCP config? +
User-level config (the default with claude mcp add) stores the server in ~/.claude/settings.json and makes it available in all projects. Project-level config stores it in a .mcp.json file in the project root, which can be committed to git for team sharing. For ContextBolt, user-level is usually best since you want your bookmarks accessible everywhere.
Can I add ContextBolt via a config file instead of the CLI? +
Yes. Add the server entry to ~/.claude/settings.json under the mcpServers key, or create a .mcp.json file in your project root. The format is the same as Claude Desktop: {"mcpServers": {"contextbolt": {"command": "npx", "args": ["-y", "@contextbolt/mcp-server"], "env": {"CONTEXTBOLT_TOKEN": "your-token"}}}}.
Do I need Node.js installed? +
Yes. The MCP server is distributed as an npm package and launched via npx. You need Node.js 18 or later installed on your system. If you already have Claude Code installed, you almost certainly have Node.js.
Can Claude Code modify my bookmarks? +
No. The MCP integration is read-only. Claude Code can search and retrieve bookmark data, but it cannot create, modify, or delete bookmarks.
How do I check if the server is connected? +
Run claude mcp list in your terminal to see all configured servers and their status. Inside an interactive session, type /mcp to see connection status for all MCP servers.