diff --git a/README.md b/README.md
index dc9be53..bfc2855 100644
--- a/README.md
+++ b/README.md
@@ -300,7 +300,7 @@ uses — just exposed at scene granularity for interactive tools.
 - [x] Scene chaining with first/last frame conditioning
 - [x] Character bank with reference images
 - [x] TTS + subtitle integration
-- [x] Auto-publishing — TikTok + YouTube (Instagram in progress)
+- [x] Auto-publishing — TikTok, YouTube, Instagram Reels + Facebook (Meta Graph API)
 - [x] Scheduler for autonomous daily generation
 - [x] Quality validation and scene regeneration
 - [x] Agentic editing — in-app agent (kimi) + MCP server over the tool registry
diff --git a/config/config.example.toml b/config/config.example.toml
index 7698e17..4b209ee 100644
--- a/config/config.example.toml
+++ b/config/config.example.toml
@@ -61,10 +61,17 @@ voice_transcript = ""  # Transcript of the voice sample
 
 [publish]
 enabled = false
-platforms = ["tiktok", "youtube", "instagram"]
+platforms = ["tiktok", "youtube", "instagram", "facebook"]
 tiktok_access_token = ""
 youtube_client_id = ""
 youtube_client_secret = ""
+# Meta (Instagram Reels + Facebook Page) share one Graph API auth. Get a
+# long-lived Page access token + link an IG Business/Creator account to the Page.
+# Instagram Reels publish by public URL; Facebook also accepts a direct upload.
+meta_page_access_token = ""
+facebook_page_id = ""
+instagram_business_account_id = ""
+meta_graph_version = "v21.0"
 
 [scheduler]
 enabled = false
diff --git a/src/flow/config.py b/src/flow/config.py
index 19bf560..27754bd 100644
--- a/src/flow/config.py
+++ b/src/flow/config.py
@@ -37,10 +37,15 @@ class TTSConfig(BaseModel):
 
 class PublishConfig(BaseModel):
     enabled: bool = False
-    platforms: list[str] = ["tiktok", "youtube", "instagram"]
+    platforms: list[str] = ["tiktok", "youtube", "instagram", "facebook"]
     tiktok_access_token: str = ""
     youtube_client_id: str = ""
     youtube_client_secret: str = ""
+    # Meta (Instagram Reels + Facebook Page) — shared Graph API auth.
+    meta_page_access_token: str = ""          # long-lived Page access token
+    facebook_page_id: str = ""                # FB Page id (Facebook publishing)
+    instagram_business_account_id: str = ""   # IG Business account id (IG publishing)
+    meta_graph_version: str = "v21.0"
 
 
 class SchedulerConfig(BaseModel):
diff --git a/src/flow/publisher.py b/src/flow/publisher.py
index 26b6b08..62e9498 100644
--- a/src/flow/publisher.py
+++ b/src/flow/publisher.py
@@ -19,8 +19,14 @@ class Publisher:
     def __init__(self, config: Config) -> None:
         self.config = config
 
-    def upload(self, video_path: Path, shot_list: ShotList) -> None:
-        """Upload video to configured platforms."""
+    def upload(self, video_path: Path, shot_list: ShotList,
+               video_url: str | None = None) -> None:
+        """Upload video to configured platforms.
+
+        ``video_url`` is a public URL for the video — **required** for Instagram
+        Reels (the Graph API ingests by URL, not binary) and optional for
+        Facebook (falls back to a binary upload when omitted).
+        """
         metadata = self._generate_metadata(shot_list)
 
         for platform in self.config.publish.platforms:
@@ -30,7 +36,9 @@ def upload(self, video_path: Path, shot_list: ShotList) -> None:
                 elif platform == "youtube":
                     self._upload_youtube(video_path, metadata)
                 elif platform == "instagram":
-                    self._upload_instagram(video_path, metadata)
+                    self._upload_instagram(video_path, metadata, video_url)
+                elif platform == "facebook":
+                    self._upload_facebook(video_path, metadata, video_url)
                 else:
                     console.print(f"  ⚠ Unknown platform: {platform}")
             except Exception as e:
@@ -118,12 +126,108 @@ def _upload_youtube(
         )
         console.print("  ✓ YouTube: uploaded")
 
+    def _caption(self, metadata: dict) -> str:
+        """Build a caption from metadata (title + description + hashtags)."""
+        parts = [metadata.get("title", ""), metadata.get("description", "")]
+        tags = " ".join(f"#{t}" for t in metadata.get("tags", []) if t)
+        if tags:
+            parts.append(tags)
+        return "\n\n".join(p for p in parts if p)[:2200]
+
     def _upload_instagram(
-        self, video_path: Path, metadata: dict
+        self, video_path: Path, metadata: dict, video_url: str | None = None
     ) -> None:
-        """Upload to Instagram Reels via Graph API.
+        """Publish an Instagram Reel via the Meta Graph API.
 
-        Requires Facebook Business account + app review.
-        Two-step: create media container → publish.
+        Three steps: create a REELS media container (by public ``video_url``) →
+        poll until processing FINISHED → publish. Needs a long-lived Page access
+        token + the linked IG Business account id.
         """
-        console.print("  → Instagram: not yet implemented")
+        import time
+
+        p = self.config.publish
+        token = p.meta_page_access_token
+        ig_id = p.instagram_business_account_id
+        if not token or not ig_id:
+            console.print(
+                "  → Instagram: not configured "
+                "(meta_page_access_token + instagram_business_account_id)"
+            )
+            return
+        if not video_url:
+            console.print(
+                "  → Instagram: Reels require a public video_url — none provided, skipping"
+            )
+            return
+
+        base = f"https://graph.facebook.com/{p.meta_graph_version}"
+        # 1. Create the media container.
+        r = httpx.post(
+            f"{base}/{ig_id}/media",
+            data={
+                "media_type": "REELS", "video_url": video_url,
+                "caption": self._caption(metadata), "access_token": token,
+            },
+            timeout=60,
+        )
+        r.raise_for_status()
+        creation_id = r.json()["id"]
+
+        # 2. Poll until the container finishes processing.
+        for _ in range(30):
+            s = httpx.get(
+                f"{base}/{creation_id}",
+                params={"fields": "status_code", "access_token": token}, timeout=30,
+            )
+            s.raise_for_status()
+            status = s.json().get("status_code")
+            if status == "FINISHED":
+                break
+            if status == "ERROR":
+                raise RuntimeError("Instagram container processing failed")
+            time.sleep(5)
+        else:
+            raise RuntimeError("Instagram container not ready in time")
+
+        # 3. Publish.
+        pub = httpx.post(
+            f"{base}/{ig_id}/media_publish",
+            data={"creation_id": creation_id, "access_token": token}, timeout=60,
+        )
+        pub.raise_for_status()
+        console.print("  ✓ Instagram: published")
+
+    def _upload_facebook(
+        self, video_path: Path, metadata: dict, video_url: str | None = None
+    ) -> None:
+        """Publish a video to a Facebook Page via the Meta Graph API.
+
+        Uses the same Page access token as Instagram. Posts by ``file_url`` when a
+        public URL is available, otherwise uploads the local file directly.
+        """
+        p = self.config.publish
+        token = p.meta_page_access_token
+        page_id = p.facebook_page_id
+        if not token or not page_id:
+            console.print(
+                "  → Facebook: not configured (meta_page_access_token + facebook_page_id)"
+            )
+            return
+
+        url = f"https://graph.facebook.com/{p.meta_graph_version}/{page_id}/videos"
+        data = {
+            "title": metadata.get("title", "")[:255],
+            "description": metadata.get("description", "")[:1000],
+            "access_token": token,
+        }
+        if video_url:
+            data["file_url"] = video_url
+            r = httpx.post(url, data=data, timeout=120)
+        else:
+            with open(video_path, "rb") as f:
+                r = httpx.post(
+                    url, data=data,
+                    files={"source": ("video.mp4", f, "video/mp4")}, timeout=600,
+                )
+        r.raise_for_status()
+        console.print("  ✓ Facebook: published")
diff --git a/tests/test_publisher.py b/tests/test_publisher.py
new file mode 100644
index 0000000..fc2484c
--- /dev/null
+++ b/tests/test_publisher.py
@@ -0,0 +1,98 @@
+"""Tests for the Meta Graph API publishers (Instagram Reels + Facebook Page)."""
+
+from pathlib import Path
+
+from flow.config import Config
+from flow.publisher import Publisher
+from flow.schemas import ShotList
+
+
+class _Resp:
+    def __init__(self, payload=None):
+        self._payload = payload or {}
+
+    def raise_for_status(self):
+        pass
+
+    def json(self):
+        return self._payload
+
+
+def _shotlist():
+    return ShotList(title="My Reel", narration="A short story.", scenes=[])
+
+
+def test_instagram_publishes_via_graph_api(monkeypatch):
+    cfg = Config()
+    cfg.publish.platforms = ["instagram"]
+    cfg.publish.meta_page_access_token = "tok"
+    cfg.publish.instagram_business_account_id = "ig123"
+    calls = []
+
+    def fake_post(url, data=None, timeout=None, **kw):
+        calls.append(url)
+        if url.endswith("/media"):
+            return _Resp({"id": "creation1"})
+        return _Resp({"id": "published1"})
+
+    def fake_get(url, params=None, timeout=None, **kw):
+        return _Resp({"status_code": "FINISHED"})
+
+    monkeypatch.setattr("flow.publisher.httpx.post", fake_post)
+    monkeypatch.setattr("flow.publisher.httpx.get", fake_get)
+
+    Publisher(cfg).upload(Path("clip.mp4"), _shotlist(), video_url="https://cdn/x.mp4")
+
+    assert any(u.endswith("/ig123/media") for u in calls)
+    assert any(u.endswith("/ig123/media_publish") for u in calls)
+
+
+def test_instagram_skips_without_public_url(monkeypatch):
+    cfg = Config()
+    cfg.publish.platforms = ["instagram"]
+    cfg.publish.meta_page_access_token = "tok"
+    cfg.publish.instagram_business_account_id = "ig123"
+    called = []
+    monkeypatch.setattr("flow.publisher.httpx.post", lambda *a, **k: called.append(1))
+
+    Publisher(cfg).upload(Path("clip.mp4"), _shotlist())  # no video_url
+    assert called == []  # honest skip, no API call
+
+
+def test_instagram_skips_when_unconfigured(monkeypatch):
+    cfg = Config()
+    cfg.publish.platforms = ["instagram"]
+    called = []
+    monkeypatch.setattr("flow.publisher.httpx.post", lambda *a, **k: called.append(1))
+
+    Publisher(cfg).upload(Path("clip.mp4"), _shotlist(), video_url="https://cdn/x.mp4")
+    assert called == []  # no token/ig id -> skip
+
+
+def test_facebook_publishes_by_file_url(monkeypatch):
+    cfg = Config()
+    cfg.publish.platforms = ["facebook"]
+    cfg.publish.meta_page_access_token = "tok"
+    cfg.publish.facebook_page_id = "page123"
+    seen = {}
+
+    def fake_post(url, data=None, timeout=None, **kw):
+        seen["url"] = url
+        seen["data"] = data
+        return _Resp({"id": "vid1"})
+
+    monkeypatch.setattr("flow.publisher.httpx.post", fake_post)
+
+    Publisher(cfg).upload(Path("clip.mp4"), _shotlist(), video_url="https://cdn/x.mp4")
+    assert seen["url"].endswith("/page123/videos")
+    assert seen["data"]["file_url"] == "https://cdn/x.mp4"
+
+
+def test_facebook_skips_when_unconfigured(monkeypatch):
+    cfg = Config()
+    cfg.publish.platforms = ["facebook"]
+    called = []
+    monkeypatch.setattr("flow.publisher.httpx.post", lambda *a, **k: called.append(1))
+
+    Publisher(cfg).upload(Path("clip.mp4"), _shotlist(), video_url="https://cdn/x.mp4")
+    assert called == []