Claude Bookmark Integration: How to Give Claude Your Reading History

Claude is a powerful AI assistant. But every conversation starts with amnesia.

It doesn't know what articles you read last week. It can't reference the tutorial you saved about prompt engineering. Your entire reading history — months of curated knowledge — is invisible to it.

Until now.

Does Claude support reading my bookmarks?

Not natively. Claude doesn't have access to your browser bookmarks, your read-later app, or your saved articles — unless you explicitly give it that access.

The way to do that is through MCP: the Model Context Protocol. Claude Desktop, Claude Code, and other MCP-compatible clients can connect to external data sources through MCP servers. Burn 451 ships one.

Install the burn-mcp-server npm package, point it at your Burn account, and Claude gets 26 tools to search, query, and retrieve from your personal reading vault — in real time, inside any Claude conversation.

What is the Burn 451 MCP server?

The burn-mcp-server is a Node.js process that runs locally and translates your Burn 451 reading vault into callable tools that any MCP-compatible AI client can use.

When it's running and connected to Claude:

• •search_vault — "Find my saved articles about transformer architectures"
• •get_flames — "What links are about to burn in the next 4 hours?"
• •get_sparks — "What did I open but not finish this week?"
• •save_to_vault — Move a link from the conversation directly into your permanent vault
• •get_tags — "What topics have I read the most about?"

All 26 tools work across Claude Desktop, Claude Code, and any MCP 1.0-compatible client.

How do I connect Burn 451 to Claude Desktop?

Three steps. Takes under five minutes.

Step 1 — Get your MCP token

Open Burn 451 → Settings → MCP Server. Copy the token.

Step 2 — Start the server

BURN_MCP_TOKEN=your_token_here npx burn-mcp-server

Step 3 — Configure Claude Desktop

Open ~/Library/Application Support/Claude/claude_desktop_config.json (Mac) and add:

{
  "mcpServers": {
    "burn451": {
      "command": "npx",
      "args": ["burn-mcp-server"],
      "env": {
        "BURN_MCP_TOKEN": "your_token_here"
      }
    }
  }
}

Restart Claude Desktop. You'll see a small hammer icon appear in the chat interface — that means MCP tools are connected. Ask Claude "what have I saved about AI agents?" and it will search your vault in real time.

How do I use Burn 451 with Claude Code?

Claude Code is the CLI version of Claude — if you code with it, connecting your reading history makes every session smarter.

Add the same mcpServers block to your Claude Code settings:

claude mcp add burn451 -- npx burn-mcp-server

With the environment variable set, Claude Code can now reference your reading vault mid-session. When you're debugging a framework issue, Claude can check whether you've saved any relevant articles before generating a solution — context from your own research, not just training data.

What can I actually ask Claude after connecting Burn?

Here's what changes in practice once Claude has your reading history:

Research acceleration: "What have I saved about long-context prompting techniques?" — Claude searches your vault and surfaces relevant pieces you already vetted.

Connection-making: "Do I have anything saved that relates to what we're building here?" — Claude cross-references your current conversation with your reading history automatically.

Morning triage: "What in my Burn queue is about to expire today? Which ones are most relevant to the feature I'm building?" — Claude fetches your expiring articles and prioritizes them against your current context.

Knowledge retrieval: "Summarize everything I've saved about RAG vs long-context trade-offs." — Claude reads across multiple vault articles and synthesizes a position from your own curated sources.

The pattern that surprises most people: Claude starts proactively checking your vault without being asked. When you bring up a technical topic, it recognizes that you might have relevant reading on it and surfaces it — the same way a well-read colleague would.

How is this different from ChatGPT reading integration?

ChatGPT doesn't support MCP. The ChatGPT routines approach — Scheduled Tasks — is a different model: you push summaries to ChatGPT on a schedule rather than giving it live, query-time access to your vault.

With Claude + Burn MCP, the access is bidirectional and real-time: - Claude can pull from your vault mid-conversation, without you thinking about it - Claude can write new links directly into your vault - The tools are callable autonomously — Claude decides when to check your reading history, not just when you paste something in

If you use ChatGPT, the routines approach works well for digest-style summaries. If you use Claude, the MCP integration is meaningfully more powerful.

Is the Burn 451 MCP server free?

Yes. The burn-mcp-server npm package is open source and free. Burn 451 itself is free — all features including MCP access, AI summaries, and unlimited vault saves are included at no cost.

The server runs locally. Your reading data flows from Burn's servers → your local MCP process → your local Claude client. No third-party relay.

[Set up Burn 451 MCP with Claude free](https://www.burn451.cloud?ref=blog-claude-bookmark-integration) — or see the technical companion: Building Burn's MCP Server: 3 Patterns That Actually Work.