ContextBolt captures every tweet, post, and article you bookmark across X, Reddit, and LinkedIn into a single semantically searchable library. The MCP endpoint exposes that library to any AI client that speaks Model Context Protocol, so Claude can pull from your saved knowledge mid-conversation without you context-switching to look things up.
This is the first MCP server purpose-built around social bookmarks. Most personal-knowledge MCPs assume you’ve manually written notes. ContextBolt assumes you’ve spent years bookmarking things you intended to remember and then forgot. The capture is automatic, the search is semantic, and the MCP makes the whole library available to your AI as a first-class memory layer.
Why this exists
Bookmarks are personal documentation, but only if you can find them again. The average X user has 2,000+ bookmarks and uses them like a black hole. Built-in search is keyword-only. There’s no tagging, no clustering, no way to ask “what did I save about pricing strategy six months ago?”.
The MCP endpoint solves the retrieval half. ContextBolt has already solved the capture half via the Chrome extension. Together they form an end-to-end memory layer for the AI clients you already use every day.
What Claude can do with it
Claude gets six MCP tools: semantic search across all bookmarks, browse your topic clusters, pull every save in a given topic, get your most recent saves, export the collection, and save a new bookmark from inside a conversation. Combine them in one prompt and Claude can do things like “find every save I have about competitor X, summarize the consensus, and bookmark this new tweet under the same topic”.
Practical patterns:
- “What did I save last quarter about onboarding flows?” Claude pulls relevant saves and summarizes them.
- “Add the top three threads I saved about AI agents to my Notion ‘AI Research’ database.” Combine ContextBolt MCP with the Notion MCP server in the same Claude Desktop session.
- “I’m writing a post on context engineering. Pull every relevant save and cite the original authors.” ContextBolt returns the structured saves with URLs intact.
Gotchas
The MCP endpoint is rate-limited at 200 requests per hour per user. That covers normal usage by an order of magnitude but heavy automation might hit the cap. If you need more, contact us and we’ll discuss.
ContextBolt only sees bookmarks captured after you install the extension. We don’t backfill historical X bookmarks because the Twitter API charges for archive access and we haven’t found a free path. If a viable archive route opens up, we’ll add it.
New connections use OAuth with the shared URL https://api.contextbolt.app/mcp. Finish the email-code sign-in and approval. The older personal URL is still supported, but it is a secret credential; keep it out of shared configs and public repositories. Rotating that personal URL does not revoke a separate OAuth connection.