Daily API surface sync (20260818)

.gitignore

@@ -16,3 +16,6 @@ src/graphify-out/
 __pycache__/
 *.egg-info/
 dist/
+
+# api-sync operational reports (internal — never commit; canonical copy in /var/log/gen-api-sync/)
+.api-sync-reports/

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "gen-mcp-server"
-version = "1.0.0"
+version = "1.2.0"
 description = "DEPRECATED — the official GEN MCP is the hosted server at https://mcp.gen.pro (Authorization: Bearer <gen_PAT>). This is the FINAL PyPI release; no further versions will be published."
 readme = "README.md"
 requires-python = ">=3.10"

src/gen_mcp_server/client.py

@@ -103,18 +103,22 @@ async def _call(
     path: str,
     body: Any | None = None,
     params: dict[str, Any] | None = None,
+    extra_headers: dict[str, str] | None = None,
 ) -> Any:
     pat = _resolve_pat()
     url = f"{base}{path}"
+    headers = {
+        "X-API-Key": pat,
+        "Content-Type": "application/json",
+        "X-Client": "gen-mcp-server",
+    }
+    if extra_headers:
+        headers.update(extra_headers)
     async with httpx.AsyncClient(timeout=120.0) as http:
         res = await http.request(
             method,
             url,
-            headers={
-                "X-API-Key": pat,
-                "Content-Type": "application/json",
-                "X-Client": "gen-mcp-server",
-            },
+            headers=headers,
             params=params,
             content=json.dumps(body) if body is not None else None,
         )
@@ -141,9 +145,14 @@ async def _call(
     return data
 
 
-async def api_call(method: str, path: str, body: Any | None = None) -> Any:
+async def api_call(
+    method: str,
+    path: str,
+    body: Any | None = None,
+    extra_headers: dict[str, str] | None = None,
+) -> Any:
     """Rails content engine — api.gen.pro/v1."""
-    return await _call(BASE_URL, method, path, body)
+    return await _call(BASE_URL, method, path, body, extra_headers=extra_headers)
 
 
 async def form_call(method: str, path: str, form: dict[str, str]) -> Any:

src/gen_mcp_server/reference.py

@@ -52,7 +52,7 @@
 # The 5-Step Journey
 
 Every user of GEN — human or AI — follows the same 5-step arc. This server's
-143 tools are organized around it. **Always know which step you're in.**
+157 tools are organized around it. **Always know which step you're in.**
 
 ```
   Step 1           Step 2            Step 3             Step 4            Step 5
@@ -261,16 +261,19 @@
 | Tool | When to use |
 |---|---|
 | `gen_get_engine` | Read the full engine (columns, rows, cells) before editing. |
+| `gen_list_rows_delta` | Poll only the rows that CHANGED since a cursor while watching a busy sheet; `row_ids` + `cursor` drive deletion reconciliation. |
 | `gen_update_cell` | Set a cell value directly (text, prompt, etc.). |
 | `gen_generate_content` | **Workhorse.** Trigger AI generation for a cell. Returns `generation_id`. |
 | `gen_generate_layer` | Trigger AI generation for a specific layer within a video cell. |
 | `gen_get_generation` | Poll generation status every 5s until `completed`. |
+| `gen_list_cell_jobs` | Fetch one cell's (or one video layer's) generation/job history without pulling the whole row. |
 | `gen_stop_generation` | Stop a running generation (refunds credits). |
 | `gen_continue_generation` | Resume a stopped generation (re-charges credits). |
 | `gen_create_layer` / `gen_update_layer` | Compose video layers inside a cell. |
-| `gen_create_row` / `gen_duplicate_row` | Add or duplicate rows for batch production. |
+| `gen_create_row` / `gen_duplicate_row` | Add or duplicate rows for batch production. Pass `idempotency_key` so a retried call replays the original result instead of creating a duplicate. |
 | `gen_import_asset_from_url` | Import a YouTube/TikTok/Instagram/direct URL as an asset. |
 | `gen_create_direct_upload` | Get a pre-signed URL for large file uploads (>50 MB). |
+| `gen_transcribe` | Standalone audio/video → text transcript with timestamps. Exactly one of `audio_url`/`video_url`/`content_resource_id`. Billed by duration. |
 | `gen_estimate_job` | Check the credit cost BEFORE triggering a paid generation. |
 
 **Generation types and models:**
@@ -354,6 +357,8 @@
 | `gen_list_scheduled_posts` | View the content calendar. |
 | `gen_get_post_status` | Poll post status — `accepted` → `publishing` → `published` or `failed`. |
 | `gen_create_recurring_job` | Set up an automated daily/weekly content job. |
+| `gen_preview_recurring_job_prompt` | Parse a recurring-job prompt into the row strategy/actions it implies — WITHOUT saving (free, read-only). Check before saving, or diff against existing actions. |
+| `gen_draft_test_recurring_job` | Test an UNSAVED recurring job once before committing to save it (configure → Test → Save). Paid-media gated. |
 
 **Render the final video:**
 
@@ -426,6 +431,8 @@
 The server flags `insufficient_credits: true` when a call fails due to low
 balance. Use `gen_get_credit_balance` to check and `gen_buy_credits` to add
 more (confirmation-gated). Use `gen_estimate_job` before expensive generations.
+`gen_list_subscriptions` shows the user's active plans across their workspaces
+(ACTIVE first).
 
 Standard error shape:
 ```json