diff --git a/proposals/4340-in-room-command-prompts.md b/proposals/4340-in-room-command-prompts.md
new file mode 100644
index 00000000000..3568a19b287
--- /dev/null
+++ b/proposals/4340-in-room-command-prompts.md
@@ -0,0 +1,131 @@
+# MSC4340: Prompts and partial commands for in room commands.
+
+When composing a command, a user may not know what to provide for as an argument to a command. In
+this situation, it is desirable for the client to be able to prompt the user with suggestions for
+the argument. 
+
+## Proposal
+
+### The promptable property on parameters
+
+Within the parameter description for a command argument, we specify a new property `promptable`
+which when given the value `true` signals to clients that they may send a _partial command_ to the
+bot to receive suggestions for the argument.
+
+### The `m.bot.command_prompt` event
+
+Consider the following command description under the context of MSC4332 and MSC4333.
+
+```jsonc
+{
+  "sigil": "!",
+  "commands": [
+    {
+      "syntax": "draupnir takedown {entity} {list} {reason}",
+      "arguments": {
+        "entity": {
+          "m.text": [{"body": "The entity to ban"}],
+        },
+        "reason": {
+          "m.text": [{"body": "The reason for the ban"}],
+        },
+        "list": {
+          "m.text": [{"body": "The policy list to ban the user on"}],
+          "promptable": true,
+        },
+      },
+      "description": {
+        "m.text": [{"body": "Ban a user in all protected rooms"}],
+      },
+    },
+  ]
+}
+```
+
+When a client is composing the argument for `list`, if the user is unable to make a decision for the
+policy list, the client can send the following `m.bot.command`:
+
+```jsonc
+{
+  "m.bot.command": {
+    "syntax": "draupnir takedown {entity} {list} {reason}",
+    "arguments": {
+      "entity": {
+        "id": "!room:example.org",
+        "via": ["second.example.org"]
+      },
+    }
+  }
+}
+```
+
+The bot will then respond with an event containing an `m.bot.command_prompt` mixin:
+
+
+```jsonc
+{
+  "body": "Please select a policy room",
+  "m.bot.command_prompt": {
+    "syntax": "draupnir takedown {entity} {list} {reason}",
+    "arguments": {
+      "entity": {
+        "id": "!room:example.org",
+        "via": ["second.example.org"]
+      },
+    },
+    "suggested_arguments": {
+      "list": {
+        "suggested": [
+          {
+            "id": "!policyroom:example.com",
+            "via": ["second.example.com"]
+          },
+          {
+            "id": "!fTjMjIzNKEsFlUIiru:neko.dev",
+            "via": ["neko.dev"]
+          }
+        ],
+        "default": {
+          "id": "!policyroom:example.com",
+          "via": ["second.example.com"]
+        }
+      }
+    }
+  }
+}
+```
+
+It is desirable that a bot can send a prompt without any prior command, so that it is possible to
+create notices which prompt action.
+
+### Client considerations
+
+Clients should allow prompts to be cancellable.
+
+## Potential issues
+
+None considered.
+
+## Alternatives
+
+### Reaction based prompts
+
+[Draupnir](https://github.com/the-draupnir-project/Draupnir) currently manages prompts manually by
+embedding a mixin in a reply to a command event with suggestsions that can be selected by sending
+a reaction that references the event.
+
+## Security considerations
+
+### Unsolicited prompt
+
+Clients need to take care to make sure that the sender of the prompt matches the bot for which the
+command is intended to be sent to. Otherwise clients could be tricked into sending commands by other
+users.
+
+## Unstable prefix
+
+`m.bot.command_prompt` -> `org.matrix.msc4340.command_prompt`
+
+## Dependencies
+
+This MSC builds on MSC4332 (which at the time of writing have not yet been accepted into the spec).