danielmiessler · brycemagera · Mar 30, 2026
diff --git a/README.md b/README.md
@@ -111,17 +111,12 @@ The website's dashboard fetches data from the MCP server at `mcp.daemon.danielmi
 
 If you want the full experience with a queryable MCP endpoint:
 
-1. The MCP server is a separate Cloudflare Worker that:
-   - Parses your `daemon.md` file
-   - Stores the data in Cloudflare KV
-   - Serves it via JSON-RPC (MCP protocol)
+The `mcp-worker/` directory contains a Cloudflare Worker that:
+- Fetches `daemon.md` from your deployed Pages site (5-minute edge cache)
+- Parses all sections and exposes them as 13 MCP tools
+- Serves JSON-RPC over HTTPS — no KV or additional infrastructure needed
 
-2. You'll need to:
-   - Create a Cloudflare Worker for your MCP endpoint
-   - Set up a KV namespace for data storage
-   - Update the dashboard component to point to your MCP URL
-
-3. The MCP server code and setup instructions will be documented separately.
+Full deployment instructions, including Claude Code integration: **[docs/mcp-setup.md](docs/mcp-setup.md)**
 
 **Note**: The static site works without the MCP component—you just won't have the live API functionality until you set up your own MCP server.
 

diff --git a/docs/mcp-setup.md b/docs/mcp-setup.md
@@ -0,0 +1,156 @@
+# MCP Server Setup
+
+This guide walks you through deploying the Daemon MCP Worker — a Cloudflare Worker that exposes your `daemon.md` data as a queryable [Model Context Protocol](https://modelcontextprotocol.io) endpoint.
+
+Once deployed, AI assistants (Claude, Cursor, etc.) can query your daemon directly.
+
+## Prerequisites
+
+- A deployed Daemon Pages site (your `daemon.md` is publicly accessible at `https://your-daemon-domain.com/daemon.md`)
+- A [Cloudflare account](https://cloudflare.com) (free tier works)
+- [Bun](https://bun.sh) installed
+- Wrangler authenticated: `bunx wrangler login`
+
+## 1. Update the daemon URL
+
+In `mcp-worker/src/index.ts`, update the first line to point to your deployed Pages site:
+
+```typescript
+const DAEMON_MD_URL = 'https://your-daemon-domain.com/daemon.md';
+```
+
+## 2. Configure the Worker
+
+Edit `mcp-worker/wrangler.toml`:
+
+```toml
+name = "yourname-daemon-mcp"   # unique Worker name in your Cloudflare account
+main = "src/index.ts"
+compatibility_date = "2025-08-20"
+workers_dev = true
+
+# Optional: add a custom domain (e.g. mcp.daemon.yourdomain.com)
+# See step 4 for custom domain setup.
+# routes = [{ pattern = "mcp.daemon.yourdomain.com", custom_domain = true }]
+
+[observability]
+enabled = false
+```
+
+> **Important:** If you add a `routes` field, it must appear **before** any `[section]` headers in the file. TOML parses keys as part of the nearest preceding section — a `routes` field inside `[observability]` will be silently ignored and your custom domain will not be registered.
+
+## 3. Deploy
+
+```bash
+cd mcp-worker
+bun install
+bun run deploy
+```
+
+Wrangler will print your Worker URL:
+
+```
+Deployed daemon-mcp triggers
+  https://yourname-daemon-mcp.your-account.workers.dev
+```
+
+Your MCP endpoint is now live at that URL. Test it:
+
+```bash
+curl -s -X POST https://yourname-daemon-mcp.your-account.workers.dev \
+  -H "Content-Type: application/json" \
+  -d '{"jsonrpc":"2.0","method":"tools/list","id":1}' | python3 -m json.tool
+```
+
+You should see a list of 13 tools, each with an `inputSchema` field.
+
+## 4. Custom domain (optional)
+
+To use a custom subdomain like `mcp.daemon.yourdomain.com`:
+
+1. Make sure `yourdomain.com` is managed by Cloudflare (nameservers pointed at Cloudflare).
+2. Uncomment and update the `routes` line in `wrangler.toml` **before** the `[observability]` section:
+
+```toml
+routes = [{ pattern = "mcp.daemon.yourdomain.com", custom_domain = true }]
+
+[observability]
+enabled = false
+```
+
+3. Run `bun run deploy` again. Cloudflare will provision the DNS record and TLS certificate automatically.
+
+DNS propagation to upstream resolvers can take up to 24 hours. You can verify it's live on Cloudflare's resolver immediately:
+
+```bash
+dig @1.1.1.1 mcp.daemon.yourdomain.com
+```
+
+## 5. Connect to Claude Code
+
+### Project-scoped (recommended)
+
+Add to `~/.claude.json` in your project root. This makes the MCP server available only when working in this project:
+
+```json
+{
+  "mcpServers": {
+    "daemon": {
+      "type": "http",
+      "url": "https://yourname-daemon-mcp.your-account.workers.dev"
+    }
+  }
+}
+```
+
+### Global (all Claude Code sessions)
+
+Add to `~/.claude/settings.json` to make it available everywhere:
+
+```json
+{
+  "mcpServers": {
+    "daemon": {
+      "type": "http",
+      "url": "https://yourname-daemon-mcp.your-account.workers.dev"
+    }
+  }
+}
+```
+
+After adding either config, run `/mcp` in Claude Code to verify the connection. You should see the server listed as connected with 13 tools available.
+
+## How it works
+
+The Worker:
+
+1. Receives JSON-RPC requests from MCP clients
+2. Fetches `daemon.md` from your Pages site (cached at the edge for 5 minutes)
+3. Parses the section-based format and routes tool calls to the appropriate section
+4. Returns results as MCP `text` content
+
+The Worker implements the [MCP Streamable HTTP transport](https://spec.modelcontextprotocol.io/specification/2025-03-26/basic/transports/#streamable-http):
+
+- `POST /` — handles all JSON-RPC requests
+- `GET /` — returns `405 Method Not Allowed` (server-push SSE not supported)
+- `OPTIONS /` — CORS preflight
+- Requests without an `id` (notifications) return `202 Accepted` with no body
+- All tool definitions include `inputSchema` — required by MCP clients for validation
+
+## Available tools
+
+| Tool | Description |
+|------|-------------|
+| `get_about` | Basic information |
+| `get_mission` | Mission statement |
+| `get_telos` | Full TELOS framework (raw text) |
+| `get_telos_structured` | TELOS as structured JSON with sub-bullets |
+| `get_current_location` | Current location |
+| `get_favorite_books` | Reading list |
+| `get_currently_reading` | Current fiction/hobby reading |
+| `get_favorite_movies` | Movie recommendations |
+| `get_favorite_podcasts` | Podcast recommendations |
+| `get_preferences` | Work style and preferences |
+| `get_predictions` | Predictions about the future |
+| `get_all` | Everything as structured JSON |
+| `get_section` | Any section by name (e.g. `ABOUT`, `TELOS`) |
diff --git a/mcp-worker/package.json b/mcp-worker/package.json
@@ -0,0 +1,12 @@
+{
+  "name": "daemon-mcp",
+  "type": "module",
+  "scripts": {
+    "dev": "wrangler dev",
+    "deploy": "wrangler deploy"
+  },
+  "devDependencies": {
+    "@cloudflare/workers-types": "^4.0.0",
+    "wrangler": "^4.0.0"
+  }
+}