Connect via MCP

Drive CacheBoost cache warming from an AI agent — Claude Desktop, Claude Code, or Cursor — using the CacheBoost MCP server.

The CacheBoost MCP server exposes cache warming and run monitoring as Model Context Protocol tools. An agent connected to the server can warm your critical URLs after a deployment and report back when the cache is hot — in natural language, without writing API calls by hand.

Endpoint

The server is a single remote Streamable HTTP endpoint:

https://api.cache-boost.com/mcp

It runs on the same host as the REST API. The protocol is session-based: your MCP client performs the initialize handshake, receives an Mcp-Session-Id header, and replays it on every subsequent tool call. Compliant clients handle this automatically.

Authentication

The MCP server uses your existing CacheBoost API key — there is no separate MCP credential. Create a key in the app under Profile → API Keys, grant it the scopes for the tools you intend to use (see the table below), then send it in the X-API-Key header:

X-API-Key: cb_live_YOUR_KEY

Authorization: Bearer cb_live_YOUR_KEY is also accepted (and takes precedence when both are present), but X-API-Key works with every MCP client, including the Claude custom connector.

A missing or invalid key is rejected with 401. The key's scopes and its optional site restriction apply exactly as they do on the REST API: a tool called without the required scope returns an explicit tool error (never a silent success), and a site outside the key's access is reported as not found. See Authentication for how to create keys and manage scopes.

Add the server to your client

CacheBoost is listed in the official MCP registry as com.cache-boost/cacheboost. If your client can browse the registry, that is the short path; everything below it is the manual equivalent.

From your client's server directory

Search for CacheBoost in your client's MCP directory, add it, and paste your cb_live_… key when it asks for X-API-Key. The endpoint and transport come from the listing — there is nothing else to configure.

If your client has no directory, register CacheBoost by hand as a remote (Streamable HTTP) MCP server with an X-API-Key header.

Claude Code (CLI)

claude mcp add --transport http cacheboost https://api.cache-boost.com/mcp \
  --header "X-API-Key: cb_live_YOUR_KEY"

Claude Desktop / Cursor

Add an entry to the MCP servers configuration:

{
  "mcpServers": {
    "cacheboost": {
      "type": "http",
      "url": "https://api.cache-boost.com/mcp",
      "headers": {
        "X-API-Key": "cb_live_YOUR_KEY"
      }
    }
  }
}

Claude (custom connector)

In Claude, open Settings → Connectors → Add custom connector and enter https://api.cache-boost.com/mcp. Then:

• Authentication: choose No authentication (CacheBoost uses an API key, not OAuth).
• Request headers: add X-API-Key with your key cb_live_YOUR_KEY. Do not use Authorization here: the connector does not send it.

Available tools

Each tool mirrors the scope of its REST counterpart. Grant the API key only the scopes for the tools you need.

warm_site keeps the same business rules as the REST API: URLs must be same-origin with the site, at most 5000 URLs per call, duplicates are removed, quota is pre-checked, and a 409 is returned if a run is already in progress for the site.

Example: warm critical URLs after a deploy

Once the server is connected, ask your agent in natural language, for example:

Warm the critical URLs of my site after this deploy and tell me when the cache is hot.

The agent calls the tools in sequence:

1. warm_site(site_id, urls) → returns { run_id, url_count, status: "pending" }.
2. get_run(run_id) (or list_warm_runs(site_id)) polled until status is done or failed, then reports the cache-hit results.