@writepanda/mcp 1.119.0 → 1.122.0

package/bin/server.mjs

@@ -391,6 +391,22 @@ const TOOLS = [
 		},
 		command: "workspace.set-brand",
 	},
+	{
+		name: "workspace_capture_brand",
+		description:
+			"Auto-populate the active workspace's brand kit from a website URL. Runs HyperFrames' capture against the site, classifies its real colors (background/ink/primary/accent), fonts (display/body), and logo, and merges them into the brand kit (hand-set fields are preserved unless capture finds a better value). ASYNC — returns { jobId }; poll job_wait. Needs network; first run downloads the HyperFrames CLI, so allow a generous timeout. Use when the user says 'use my brand', 'pull my brand from <site>', or onboards a client by URL.",
+		inputSchema: {
+			type: "object",
+			properties: {
+				url: {
+					type: "string",
+					description: "The brand's website URL (e.g. https://acme.com). https:// is assumed if omitted.",
+				},
+			},
+			required: ["url"],
+		},
+		command: "workspace.capture-brand",
+	},
 	{
 		name: "workspace_get_project_defaults",
 		description:
@@ -2623,7 +2639,7 @@ const TOOLS = [
 	{
 		name: "caption_set_template",
 		description:
-			"Pick a caption template: classic, modern, minimal, bold, spotlight, boxed, neon, colored, or editorial. `editorial` is a magazine-emphasis style: the word being spoken right now renders large (and takes an accent color) while the rest of the line shrinks, so one big word sweeps across the line in time with the speech.",
+			"Pick a caption template. Static styles: classic, modern, minimal, bold, spotlight, boxed, neon, colored, editorial (`editorial` = magazine emphasis: the spoken word renders large + accent-colored while the rest of the line shrinks). ANIMATED, transcript-driven styles (each word animates as it's spoken, in preview AND export): kineticSlam (words slam in), clipWipe (left-to-right wipe per word), gradientPop (gradient text, elastic pop), matrixDecode (character scramble that resolves), glitchRgb (RGB chromatic split), blendDifference (auto-inverts over any footage).",
 		inputSchema: {
 			type: "object",
 			properties: {
@@ -2641,6 +2657,12 @@ const TOOLS = [
 						"neon",
 						"colored",
 						"editorial",
+						"kineticSlam",
+						"clipWipe",
+						"gradientPop",
+						"matrixDecode",
+						"glitchRgb",
+						"blendDifference",
 					],
 				},
 				expectedRevision: { type: "number" },
@@ -2795,7 +2817,7 @@ const TOOLS = [
 	{
 		name: "media_generate_narration",
 		description:
-			"Generate a voiceover / narration audio clip. Project-agnostic; writes audio to <userData>/narration/ and returns { audioPath, durationMs, model, voice, predictionId }. Canonical workflow: media_generate_narration → project_add_audio (place at startMs, pass the returned durationMs so the overlay is sized to the speech). DEFAULT is the LOCAL, on-device Kokoro engine (English) — no API key, no cloud, runs on the user's machine; the model auto-downloads (~330 MB) on first use. Kokoro voices (pass in `voice`): af_heart (default), af_bella, am_michael, bf_emma, bm_george, etc. (a*=American, b*=British; f=female, m=male). To use CLOUD TTS instead (more voices/languages, needs the user's Replicate key), set `model` to a Replicate model: 'elevenlabs-v3' (most expressive — embed inline tags like [excited]/[whispers]) | 'gemini-flash-tts' (30 voices, multilingual, 'style' sets tone) | 'minimax-turbo' (fast, 'style' maps to an emotion, 'speed' 0.5-2). Cloud voices — gemini Kore/Puck/Charon; minimax Friendly_Person/Wise_Woman. Omit `model` to honour the workspace default (system_get_narration_engine).",
+			"Generate a voiceover / narration audio clip. Project-agnostic; writes audio to <userData>/narration/ and returns { audioPath, durationMs, model, voice, predictionId }. Canonical workflow: media_generate_narration → project_add_audio (place at startMs, pass the returned durationMs so the overlay is sized to the speech). DEFAULT is the LOCAL, on-device Kokoro engine (English) — no API key, no cloud, runs on the user's machine; the model auto-downloads (~330 MB) on first use. Kokoro voices (pass in `voice`): af_heart (default), af_bella, am_michael, bf_emma, bm_george, etc. (a*=American, b*=British; f=female, m=male). To use CLOUD TTS instead (more voices/languages, needs the user's Replicate key), set `model` to a Replicate model: 'elevenlabs-v3' (most expressive — embed inline tags like [excited]/[whispers]; stock voices only) | 'elevenlabs-direct' (the user's OWN ElevenLabs account — CLONED voices; needs the ElevenLabs key from Settings → Integrations; pass the voice NAME or voice_id) | 'gemini-flash-tts' (30 voices, multilingual, 'style' sets tone) | 'minimax-turbo' (fast, 'style' maps to an emotion, 'speed' 0.5-2). Cloud voices — gemini Kore/Puck/Charon; minimax Friendly_Person/Wise_Woman. Omit `model` to honour the workspace default (system_get_narration_engine).",
 		inputSchema: {
 			type: "object",
 			properties: {
@@ -2807,7 +2829,7 @@ const TOOLS = [
 				model: {
 					type: "string",
 					description:
-						"Engine/model selector. Omit for the workspace default (local Kokoro). 'kokoro-local' forces on-device English. Cloud (needs Replicate key): 'elevenlabs-v3' | 'gemini-flash-tts' | 'minimax-turbo'.",
+						"Engine/model selector. Omit for the workspace default (local Kokoro). 'kokoro-local' forces on-device English. Cloud (needs Replicate key): 'elevenlabs-v3' | 'gemini-flash-tts' | 'minimax-turbo'. Direct: 'elevenlabs-direct' (own ElevenLabs key; cloned voices).",
 				},
 				voice: {
 					type: "string",

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@writepanda/mcp",
-	"version": "1.119.0",
+	"version": "1.122.0",
 	"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",