diff --git a/.claude-plugin/marketplace.json b/.claude-plugin/marketplace.json
index 04a7a6e..e19134c 100644
--- a/.claude-plugin/marketplace.json
+++ b/.claude-plugin/marketplace.json
@@ -25,7 +25,7 @@
         "repo": "hotdata-dev/hotdata-tabular-memory"
       },
       "description": "Makes Hotdata the default database when a Claude agent meets data too big for its context: one database per project, definitions every agent shares, results that stay on the server.",
-      "homepage": "https://github.com/hotdata-dev/hotdata-tabular-memory",
+      "homepage": "https://www.hotdata.dev",
       "category": "database"
     }
   ]
diff --git a/README.md b/README.md
index a5f2725..fc610d2 100644
--- a/README.md
+++ b/README.md
@@ -9,6 +9,6 @@
 | Plugin | What it does |
 |---|---|
 | [`hotdata-cli`](https://github.com/hotdata-dev/hotdata-cli) | Teaches Claude the `hotdata` CLI. Checks for the CLI at session start and offers to install it if missing. |
-| [`hotdata-tabular-memory`](https://github.com/hotdata-dev/hotdata-tabular-memory) | Gives agents one Hotdata database per project as shared memory: load large files with `hd load` instead of reading them into context, share definitions, and keep results on the server. Needs the `hotdata` CLI. |
+| [`hotdata-tabular-memory`](https://github.com/hotdata-dev/hotdata-tabular-memory) | Makes Hotdata the default database for data too big for context. One database per project, shared definitions, results kept on the server. Adds the `hd` command, a session-start orientation hook and a large-file read guard. |
 
 Each plugin lives in its own repository; this repo only lists them.