diff --git a/apps/web/public/.well-known/openmcp.json b/apps/web/public/.well-known/openmcp.json
index c3a7b46..e93a3f1 100644
--- a/apps/web/public/.well-known/openmcp.json
+++ b/apps/web/public/.well-known/openmcp.json
@@ -26,6 +26,7 @@
     "list_authors",
     "get_author",
     "directory_stats",
+    "trending_topics",
     "submit_feed",
     "get_openprofile",
     "update_openprofile",
diff --git a/apps/web/src/lib/llms.js b/apps/web/src/lib/llms.js
index 9a7067c..4e33174 100644
--- a/apps/web/src/lib/llms.js
+++ b/apps/web/src/lib/llms.js
@@ -143,7 +143,7 @@ export async function llmsTxt(opts = {}) {
     '',
     'The same capabilities as the endpoints below, as tools an agent can call:',
     'search, list_feeds, get_feed, list_topics, get_topic, topic_posts, read_post,',
-    'random_feed, directory_stats and submit_feed. Point an MCP client at',
+    'random_feed, directory_stats, trending_topics and submit_feed. Point an MCP client at',
     `\`${base}/mcp\` — it answers both the current protocol revision and the`,
     'older handshake-based ones. Documentation for humans is at the same URL in a',
     `browser: ${base}/mcp`,
diff --git a/apps/web/src/lib/mcp/tools.js b/apps/web/src/lib/mcp/tools.js
index f8e25aa..ca352bb 100644
--- a/apps/web/src/lib/mcp/tools.js
+++ b/apps/web/src/lib/mcp/tools.js
@@ -1,4 +1,4 @@
-import { q, newId, authors as people, webrings } from '@rssamplifier/db';
+import { q, newId, authors as people, trends, webrings } from '@rssamplifier/db';
 import { topicSlug } from '@rssamplifier/feed';
 import { submitCatalogue, hashIp, EXPRESS_MAX } from '@rssamplifier/ingest';
 import { submitFeedTool } from '@profullstack/submit-feed/core';
@@ -652,6 +652,27 @@ export const TOOLS = [
     },
   },
 
+  {
+    name: 'trending_topics',
+    title: 'What the open web is writing about more than last week',
+    description:
+      'Subjects English blogs and news sites used in headlines more than the week before, ranked. `count` is feeds, not posts, in a one in four sample, so it measures movement rather than coverage, and `prior_count` is the same figure for the week before. Updated daily. An empty list means the rollup has not got two days to compare yet.',
+    inputSchema: {
+      type: 'object',
+      properties: {
+        window: { type: 'integer', minimum: 1, maximum: 14, description: 'Days per side of the comparison. Default 7.' },
+        limit: { type: 'integer', minimum: 1, maximum: 200, description: 'Default 50.' },
+      },
+    },
+    annotations: { readOnlyHint: true, openWorldHint: true },
+    async run({ window = 7, limit = 50 } = {}) {
+      const size = Math.min(14, Math.max(1, Math.round(Number(window)) || 7));
+      const top = Math.min(200, Math.max(1, Math.round(Number(limit)) || 50));
+      const result = await trends.trendingTerms(db(), { window: size, limit: top });
+      return { window_days: size, generated_at: result.generatedAt, documents: result.feeds, topics: result.topics };
+    },
+  },
+
   {
     name: 'list_rings',
     title: 'List the webrings this site hosts',