diff --git a/README.md b/README.md
index 96749a0..2e1a97d 100644
--- a/README.md
+++ b/README.md
@@ -204,6 +204,8 @@ Blossom is a protocol for storing blobs (files) on media servers using your Nost
 2. Click **Publish to Blossom (with header)** (or run **Nostr: Publish to Blossom (with header)**).
 3. The front-matter block is stripped, the markdown body is uploaded as a blob, and a kind 1063 event is built from the descriptor + front-matter (title, author, language, year, cover, tags) and published to your relays.
 
+> **Markdown MIME handling:** markdown blobs are uploaded as `text/plain` so browsers render them inline (no mainstream browser displays `text/markdown`, and servers send `X-Content-Type-Options: nosniff` which forbids guessing). The kind 1063 event's `m` tag still advertises `text/markdown`, so Nostr clients that render markdown can detect the file's true type.
+
 **Blossom servers**:
 
 - The **Blossom Servers** section shows your servers with checkboxes — exactly like the Relays section. Every checked server receives the upload.
diff --git a/package.json b/package.json
index 7be933c..a8d1276 100644
--- a/package.json
+++ b/package.json
@@ -2,7 +2,7 @@
   "name": "nostr-publish",
   "displayName": "Nostr Publish",
   "description": "Publish files to Nostr from Codium/VSCode. Phase 0: raw unsigned event JSON. Phase 1: long-form notes (.md). Phase 2: signer (Rust n_signer port) signing backend.",
-  "version": "0.1.1",
+  "version": "0.1.2",
   "publisher": "laantungir",
   "license": "MIT",
   "repository": {
diff --git a/src/extension.ts b/src/extension.ts
index 08a79f7..cb9e7d7 100644
--- a/src/extension.ts
+++ b/src/extension.ts
@@ -1869,6 +1869,12 @@ function resolveBlossomServers(): string[] {
  * 3. Uploads to all checked Blossom servers concurrently.
  * 4. Builds a kind 1063 NIP-94 event from the first successful descriptor.
  * 5. Signs and publishes the 1063 event to the enabled relays.
+ *
+ * Markdown bodies are uploaded as `text/plain` (browsers render text/plain
+ * inline; no mainstream browser displays text/markdown, and servers send
+ * `X-Content-Type-Options: nosniff` which forbids guessing). The kind 1063
+ * event's `m` tag still advertises `text/markdown` so Nostr clients that
+ * render markdown can detect the file's true semantic type.
  */
 async function uploadAndPublishMetadata(
   state: StateService,
@@ -1877,6 +1883,12 @@ async function uploadAndPublishMetadata(
   body: Buffer,
   mime: string,
 ): Promise<void> {
+  // Hybrid MIME handling: markdown uploads go out as text/plain for
+  // browser-renderability; the semantic type is preserved for the m tag.
+  const isMarkdown = mime === "text/markdown";
+  const uploadMime = isMarkdown ? "text/plain" : mime;
+  const semanticMime = isMarkdown ? "text/markdown" : undefined;
+
   // Require sign-in.
   let signer = getSigner(state, config);
   if (!signer) {
@@ -1938,7 +1950,7 @@ async function uploadAndPublishMetadata(
       results = await uploadBlobToMany(
         {
           body,
-          contentType: mime,
+          contentType: uploadMime,
           sha256,
           signer,
           authTokenTtlSec: config.blossomAuthTokenTtlSec,
@@ -1966,7 +1978,12 @@ async function uploadAndPublishMetadata(
 
   // Build the kind 1063 event from the first successful descriptor.
   const descriptor: BlobDescriptor = ok[0].descriptor!;
-  const template = buildFileMetadataEvent({ descriptor, meta, publishedAt });
+  const template = buildFileMetadataEvent({
+    descriptor,
+    meta,
+    publishedAt,
+    mimeType: semanticMime,
+  });
 
   // Validate.
   const validation = validateEventTemplate(template);
diff --git a/src/nostr/nip94.ts b/src/nostr/nip94.ts
index cf70859..fc13bec 100644
--- a/src/nostr/nip94.ts
+++ b/src/nostr/nip94.ts
@@ -13,6 +13,17 @@ export interface FileMetadataInput {
   meta: BlossomFrontMatter;
   /** First-publish timestamp (unix seconds). Reused on updates. */
   publishedAt: number;
+  /**
+   * Override for the `m` tag (the file's semantic MIME type).
+   *
+   * Markdown blobs are uploaded as `text/plain` so browsers render them
+   * inline (no mainstream browser displays `text/markdown`, and
+   * `X-Content-Type-Options: nosniff` forbids guessing). The server then
+   * reports `type: text/plain` in the descriptor. This override keeps the
+   * Nostr `m` tag advertising the true semantic type (`text/markdown`) so
+   * clients that render markdown can still detect it.
+   */
+  mimeType?: string;
 }
 
 /**
@@ -47,10 +58,21 @@ export function buildFileMetadataEvent(input: FileMetadataInput): EventTemplate
 
   // Ensure the core tags exist even if nip94 omitted them.
   ensureTag(tags, "url", descriptor.url);
-  ensureTag(tags, "m", descriptor.type);
+  ensureTag(tags, "m", input.mimeType ?? descriptor.type);
   ensureTag(tags, "x", descriptor.sha256);
   ensureTag(tags, "size", String(descriptor.size));
 
+  // When an explicit mimeType override is given, replace any server-provided
+  // `m` tag (BUD-08 nip94 tags carry the upload Content-Type, which for
+  // markdown is deliberately text/plain — see FileMetadataInput.mimeType).
+  if (input.mimeType) {
+    for (const t of tags) {
+      if (t[0] === "m") {
+        t[1] = input.mimeType;
+      }
+    }
+  }
+
   // Descriptive metadata from front-matter.
   tags.push(["title", meta.title]);
   if (meta.summary) {