@writepanda/mcp 1.51.0 → 1.56.0

package/bin/server.mjs

@@ -599,7 +599,7 @@ const TOOLS = [
 	{
 		name: "project_add_motion_graphic",
 		description:
-			"Drop a motion-graphic MP4/WebM onto the timeline. PREFERRED: pass `fromJob` (the jobId from motion_render_html / motion_generate / motion_concat) — the server resolves the outputPath internally, so you never handle the path string at all. Fallback: pass `file` (an absolute path) only for external/uploaded files. Optional SFX plays when it appears.",
+			"Drop a motion-graphic MP4/WebM onto the timeline. PREFERRED: pass `fromJob` (the jobId from motion_render_html / motion_generate / motion_concat) — the server resolves the outputPath internally, so you never handle the path string at all. Fallback: pass `file` (an absolute path) only for external/uploaded files. **Defaults to a `bundled:sound/mouse-click` SFX as the graphic appears** — pass `soundUrl: null` to make it silent, or any other `bundled:sound/<id>` to swap the sound.",
 		inputSchema: {
 			type: "object",
 			properties: {
@@ -618,13 +618,13 @@ const TOOLS = [
 				durationMs: { type: "number" },
 				atMs: { type: "number", description: "Default: end of timeline" },
 				soundUrl: {
-					type: "string",
+					type: ["string", "null"],
 					description:
-						"Optional SFX played when the motion graphic appears. Bundled id (bundled:sound/<id>) like bundled:sound/message-pop, bundled:sound/swoosh-fast, etc. — use asset_list_sounds to discover. No default.",
+						"SFX played when the motion graphic appears. Defaults to bundled:sound/mouse-click. Pass any other bundled id (bundled:sound/<id> — bundled:sound/swoosh-fast, bundled:sound/message-pop, etc.; use asset_list_sounds to discover) to swap the sound, or pass null explicitly to make the graphic silent.",
 				},
 				soundVolume: {
 					type: "number",
-					description: "Sound volume 0–1. Default 1 when soundUrl is set.",
+					description: "Sound volume 0–1. Default 1. Ignored when soundUrl is null.",
 				},
 				anchorSourceMs: {
 					type: "number",
@@ -690,13 +690,13 @@ const TOOLS = [
 					description: "Slide-in/out window at each edge in ms. Default 320.",
 				},
 				soundUrl: {
-					type: "string",
+					type: ["string", "null"],
 					description:
-						"Optional SFX when the segment appears. Bundled id (bundled:sound/<id>), absolute path, or file:// URL. Use asset_list_sounds to discover.",
+						"SFX when the segment appears. Defaults to bundled:sound/mouse-click (inherited from project.add-motion-graphic). Pass any other bundled:sound/<id> to swap, or pass null explicitly to make the segment silent.",
 				},
 				soundVolume: {
 					type: "number",
-					description: "Sound volume 0–1. Default 1 when soundUrl is set.",
+					description: "Sound volume 0–1. Default 1. Ignored when soundUrl is null.",
 				},
 				anchorSourceMs: {
 					type: "number",
@@ -1125,7 +1125,7 @@ const TOOLS = [
 	{
 		name: "project_remove_region",
 		description:
-			"Delete an existing region (zoom, trim, speed, annotation, fx, lower-third, overlay, audio-overlay) by its id. Use project_read to find region ids — visual regions live under editor.*Regions; audio overlays live under audioOverlays[]. Returns the updated project.",
+			"Delete an existing region (zoom, trim, speed, annotation, fx, lower-third, overlay, clip-transform, audio-overlay) by its id. Use project_read to find region ids — visual regions live under editor.*Regions; audio overlays live under audioOverlays[]. If the region belongs to a link group (designed-segment pair: a clip-transform parking the camera into one half + the panel motion-graphic on the other half), deleting either half also removes its peer so the pair stays coherent. Returns the updated project.",
 		inputSchema: {
 			type: "object",
 			properties: {
@@ -1134,7 +1134,7 @@ const TOOLS = [
 				regionType: {
 					type: "string",
 					description:
-						"zoom | trim | speed | annotation | fx | lower-third | overlay | audio-overlay",
+						"zoom | trim | speed | annotation | fx | lower-third | overlay | clip-transform | audio-overlay",
 				},
 				regionId: { type: "string", description: "id of the region to delete" },
 				expectedRevision: { type: "number" },
@@ -1165,7 +1165,7 @@ const TOOLS = [
 	{
 		name: "project_update_region",
 		description:
-			"Patch fields on an existing region without replacing it. Accepts any subset of the region's own properties — only the supplied fields are changed, everything else is preserved. Works on zoom (depth, focusX, focusY), trim (startMs, endMs), speed (startMs, endMs, speed), lower-third (content, subtitle, atMs, durationMs, designType, accentColor, textColor, backgroundColor, backgroundRadius, fontSize, fontFamily), fx (atMs, durationMs), annotation, overlay, and audio-overlay (startMs, endMs, sourceStartMs, volume — lets you drag-trim an audio overlay without removing it).",
+			"Patch fields on an existing region without replacing it. Accepts any subset of the region's own properties — only the supplied fields are changed, everything else is preserved. Works on zoom (depth, focusX, focusY), trim (startMs, endMs), speed (startMs, endMs, speed), lower-third (content, subtitle, atMs, durationMs, designType, accentColor, textColor, backgroundColor, backgroundRadius, fontSize, fontFamily), fx (atMs, durationMs), annotation, overlay, clip-transform (startMs, endMs, preset, transitionMs), and audio-overlay (startMs, endMs, sourceStartMs, volume — lets you drag-trim an audio overlay without removing it). Time shifts on a link-grouped region (e.g. one half of a designed segment) propagate to every peer in the group so the pair stays coherent.",
 		inputSchema: {
 			type: "object",
 			properties: {
@@ -1174,7 +1174,7 @@ const TOOLS = [
 				regionType: {
 					type: "string",
 					description:
-						"zoom | trim | speed | annotation | fx | lower-third | overlay | audio-overlay",
+						"zoom | trim | speed | annotation | fx | lower-third | overlay | clip-transform | audio-overlay",
 				},
 				regionId: { type: "string", description: "id of the region to update" },
 				patch: {
@@ -1193,7 +1193,7 @@ const TOOLS = [
 	{
 		name: "project_set_webcam_layout",
 		description:
-			"Set the webcam layout preset and optional position/scale for the current project. Use 'none' to hide the webcam overlay entirely.",
+			"Set the webcam layout preset and optional position/scale for the current project. Use 'none' to hide the webcam overlay entirely. Use 'podcast' for a two-speaker composite (host + guest as co-equal tiles).",
 		inputSchema: {
 			type: "object",
 			properties: {
@@ -1202,7 +1202,7 @@ const TOOLS = [
 				preset: {
 					type: "string",
 					description:
-						"picture-in-picture | vertical-stack | side-by-side | none. 'none' hides the webcam.",
+						"picture-in-picture | vertical-stack | side-by-side | podcast | none. 'podcast' shows two co-equal speaker tiles; 'none' hides the webcam.",
 				},
 				cx: {
 					type: "number",
@@ -1443,20 +1443,40 @@ const TOOLS = [
 	{
 		name: "project_set_clip_kind",
 		description:
-			"Stamp a clip's capture origin authoritatively: kind = camera | screen | upload. project.read returns each clip's kind (your visual-strategy signal: camera → designed segments + emphasis zooms; screen → cursor zooms). On pre-v1.28 projects kind wasn't recorded, so project.read INFERS it (flagged kindInferred:true) — and inference can be wrong. Use this to correct/lock it. `camera` = talking-head; `screen` = screen recording (maybe with webcam PiP); `upload` = imported file.",
+			"Stamp a clip's capture origin authoritatively: kind = camera | screen | upload | podcast. project.read returns each clip's kind (your visual-strategy signal: camera → designed segments + emphasis zooms; screen → cursor zooms). On pre-v1.28 projects kind wasn't recorded, so project.read INFERS it (flagged kindInferred:true) — and inference can be wrong. Use this to correct/lock it. `camera` = talking-head; `screen` = screen recording (maybe with webcam PiP); `upload` = imported file; `podcast` = two-speaker composite (host=mediaPath, guest=webcamPath).",
 		inputSchema: {
 			type: "object",
 			properties: {
 				id: { type: "string" },
 				path: { type: "string" },
 				clipId: { type: "string", description: "Clip id from project_read mainTrack.clips[n].id" },
-				kind: { type: "string", enum: ["camera", "screen", "upload"] },
+				kind: { type: "string", enum: ["camera", "screen", "upload", "podcast"] },
 				expectedRevision: { type: "number" },
 			},
 			required: ["clipId", "kind"],
 		},
 		command: "project.set-clip-kind",
 	},
+	{
+		name: "project_set_webcam_offset",
+		description:
+			"Fine-tune a podcast composite's guest/host sync. offsetMs shifts the guest (webcamPath) relative to the host: positive delays the guest, negative advances it, 0 clears it (±10000). Honored in preview + export — the guest VIDEO and AUDIO shift together. Use when the two speakers look or sound slightly out of sync.",
+		inputSchema: {
+			type: "object",
+			properties: {
+				id: { type: "string" },
+				path: { type: "string" },
+				clipId: { type: "string", description: "Clip id of the podcast composite" },
+				offsetMs: {
+					type: "number",
+					description: "Guest time nudge in ms (±10000). Positive delays guest; 0 clears.",
+				},
+				expectedRevision: { type: "number" },
+			},
+			required: ["clipId", "offsetMs"],
+		},
+		command: "project.set-webcam-offset",
+	},
 
 	// ── transcript-based editing ────────────────────────────────────
 	{
@@ -1509,12 +1529,17 @@ const TOOLS = [
 	{
 		name: "transcript_remove_fillers",
 		description:
-			"Auto-detect filler words ('um', 'uh', 'like', 'you know', ...) plus immediately-repeated words and bulk-trim them.",
+			"Auto-detect filler words and bulk-trim them. By DEFAULT removes only vocalised pauses (um, uh, uhm, umm, hmm, hm) — sounds with no lexical meaning, so removing every occurrence never breaks a sentence. Pass aggressive=true to ALSO catch lexical-word fillers (like, you know, i mean, sort of, kind of); these can false-positive on legitimate uses ('I like this template' loses 'like'), so the agent should only opt in when the user explicitly asks for a thorough cleanup. Also strips immediately-repeated words ('the the') unless includeRepeats=false.",
 		inputSchema: {
 			type: "object",
 			properties: {
 				id: { type: "string" },
 				path: { type: "string" },
+				aggressive: {
+					type: "boolean",
+					description:
+						"Include lexical-word fillers (like, you know, i mean, sort of, kind of) in addition to vocalised pauses. False positives possible on legitimate uses. Default false.",
+				},
 				includeRepeats: { type: "boolean", description: "Default true" },
 				expectedRevision: { type: "number" },
 			},
@@ -2225,7 +2250,7 @@ const TOOLS = [
 const server = new Server(
 	{
 		name: "pandastudio",
-		version: "1.19.0",
+		version: "1.56.0",
 	},
 	{
 		capabilities: {

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@writepanda/mcp",
-	"version": "1.51.0",
+	"version": "1.56.0",
 	"description": "Model Context Protocol server for PandaStudio. Exposes the desktop video editor's automation surface to Cursor, Continue, Cline, Claude Desktop, and any MCP-compliant client.",
 	"keywords": [
 		"pandastudio",