@vibeframe/mcp-server 0.75.0 → 0.76.0

package/README.md

@@ -8,7 +8,7 @@ Confirmed MCP hosts today: **Claude Desktop**, **Cursor**, **OpenCode**, and **C
 
 | Surface | Package | How you call it |
 |---------|---------|-----------------|
-| MCP host (Claude Desktop / Cursor / OpenCode / Claude Code) | `@vibeframe/mcp-server` *(this)* | host calls tool by name, for example `mcp__vibeframe__scene_build({...})` |
+| MCP host (Claude Desktop / Cursor / OpenCode / Claude Code) | `@vibeframe/mcp-server` *(this)* | host calls tool by name, for example `mcp__vibeframe__build({...})` |
 | Shell / scripts (any agent host: Codex / Aider / Gemini CLI / etc.) | `@vibeframe/cli` | `vibe init my-video && vibe build my-video && vibe render my-video` |
 | Standalone agent REPL | `@vibeframe/cli` (`vibe agent`) | natural language -> CLI calls |
 
@@ -74,7 +74,7 @@ claude mcp add vibeframe -- npx -y @vibeframe/mcp-server
 Once connected, your MCP host can resolve prompts like these into typed tool calls:
 
 > "Scaffold a 12-second Swiss-Pulse promo project, three beats, and render it"
-> *→ `scene_init` + 3× `scene_add` + `scene_render`*
+> *→ `init` + 3× `scene_add` + `render`*
 
 > "Generate a cinematic backdrop image, animate it for 5 seconds, add narration"
 > *→ `generate_image` + `generate_motion` + `generate_speech`*
@@ -84,20 +84,25 @@ Once connected, your MCP host can resolve prompts like these into typed tool cal
 
 ## Available Tools
 
-Tool names are MCP-side. Your host typically prefixes them (e.g. Claude shows them as `mcp__vibeframe__scene_init`). Each one wraps the same engine call as the matching `vibe` CLI subcommand.
+Tool names are MCP-side. Your host typically prefixes them (e.g. Claude shows them as `mcp__vibeframe__init`). Each one wraps the same engine call as the matching `vibe` CLI subcommand.
 
-### Scene authoring
+### Project flow (top-level)
+
+| Tool | Description |
+|------|-------------|
+| `init` | Scaffold a video project with `STORYBOARD.md` + `DESIGN.md` |
+| `build` | Build a storyboard project: narration TTS, image assets, scene HTML composition |
+| `render` | Deterministic Hyperframes render → MP4/WebM/MOV |
+
+### Scene authoring (lower-level)
 
 | Tool | Description |
 |------|-------------|
-| `scene_init` | Low-level scene project scaffold with `STORYBOARD.md` + `DESIGN.md` |
 | `scene_styles` | List the 8 vendored visual identities (Swiss Pulse, Data Drift, …) or fetch one |
 | `scene_add` | Append a beat (narration + backdrop + composed HTML) |
 | `scene_install_skill` | Install the Hyperframes skill bundle into a scene project |
 | `scene_lint` | Validate composition HTML against the visual identity |
-| `scene_render` | Deterministic Hyperframes render → MP4 |
 | `scene_compose_prompts` | Emit the per-beat compose plan without making an LLM call |
-| `scene_build` | Build a storyboard project from `STORYBOARD.md` cues with narration, image assets, composition, and render steps |
 
 ### Generation (13)
 
@@ -139,7 +144,7 @@ Tool names are MCP-side. Your host typically prefixes them (e.g. Claude shows th
 | Tool | Description |
 |------|-------------|
 | `audio_dub` | AI voice dubbing (ElevenLabs) |
-| `audio_voice_clone` | Voice clone from sample |
+| `audio_clone_voice` | Voice clone from sample |
 | `audio_isolate` | Vocal / background isolation |
 | `audio_duck` | Auto-duck BGM under speech |
 | `audio_transcribe` | Transcript with word-level timing (Whisper) |
@@ -152,14 +157,14 @@ Tool names are MCP-side. Your host typically prefixes them (e.g. Claude shows th
 | `detect_scenes` | Find shot boundaries |
 | `detect_beats` | Find music beats |
 
-### Analysis (4)
+### Inspection (4)
 
 | Tool | Description |
 |------|-------------|
-| `analyze_media` | Unified image / video / YouTube analysis (Gemini) |
-| `analyze_video` | Temporal video understanding (Gemini) |
-| `analyze_review` | AI video review + auto-fix suggestions |
-| `analyze_suggest` | Natural-language project edit suggestions (Gemini); optional auto-apply |
+| `inspect_media` | Unified image / video / YouTube analysis (Gemini) |
+| `inspect_video` | Temporal video understanding (Gemini) |
+| `inspect_review` | AI video review + auto-fix suggestions |
+| `inspect_suggest` | Natural-language project edit suggestions (Gemini); optional auto-apply |
 
 ### Timeline (10)
 
@@ -179,14 +184,14 @@ Tool names are MCP-side. Your host typically prefixes them (e.g. Claude shows th
 | `project_create` / `project_info` | `.vibe.json` lifecycle |
 | `export_video` | Export project to MP4/WebM/MOV via FFmpeg |
 
-### Pipelines (4)
+### Remix & pipelines (4)
 
 | Tool | Description |
 |------|-------------|
-| `pipeline_run` | Execute a multi-stage YAML pipeline |
-| `pipeline_highlights` | Long-form → highlight clips |
-| `pipeline_auto_shorts` | Long-form → vertical shorts |
-| `pipeline_regenerate_scene` | Re-render a single scene against an existing storyboard.{yaml,json} |
+| `run` | Execute a multi-stage YAML pipeline (`vibe run pipeline.yaml`) |
+| `remix_highlights` | Long-form → highlight clips |
+| `remix_auto_shorts` | Long-form → vertical shorts |
+| `remix_regenerate_scene` | Re-render a single scene against an existing storyboard.{yaml,json} |
 
 ### Walkthrough (1)
 
@@ -225,7 +230,7 @@ API keys are read from the host's environment (`~/.zshrc`, MCP config `env` bloc
 | Variable | Used by |
 |----------|---------|
 | `OPENAI_API_KEY` | gpt-image-2, Whisper, GPT |
-| `ANTHROPIC_API_KEY` | Claude (translate-srt, highlights, scene_build compose pipeline) |
+| `ANTHROPIC_API_KEY` | Claude (translate-srt, highlights, build compose pipeline) |
 | `GOOGLE_API_KEY` | Gemini (analyze, review, silence-cut, narrate) |
 | `ELEVENLABS_API_KEY` | TTS, voice-clone, dubbing, SFX |
 | `XAI_API_KEY` | Grok |

package/dist/index.js

@@ -462655,7 +462655,7 @@ var sceneStylesTool = defineTool({
   name: "scene_styles",
   category: "scene",
   cost: "free",
-  description: "List the 8 vendored visual identities available for `scene_init --visual-style` (Swiss Pulse, Data Drift, \u2026) or, when `name` is provided, return the full DESIGN.md hard-gate body for one style. The DESIGN.md content is what the LLM uses as a non-negotiable visual rulebook during compose-scenes-with-skills.",
+  description: "List the 8 vendored visual identities available for `init --visual-style` (Swiss Pulse, Data Drift, \u2026) or, when `name` is provided, return the full DESIGN.md hard-gate body for one style. The DESIGN.md content is what the LLM uses as a non-negotiable visual rulebook during compose-scenes-with-skills.",
   schema: sceneStylesSchema,
   async execute(args) {
     if (args.name) {
@@ -462715,7 +462715,7 @@ var sceneInitSchema = z.object({
   duration: z.number().optional().describe("Default root composition duration in seconds. Default 10.")
 });
 var sceneInitTool = defineTool({
-  name: "scene_init",
+  name: "init",
   category: "scene",
   cost: "free",
   description: "Scaffold a new VibeFrame video scene project. Supports minimal, agent, and full profiles; full includes the current HTML render backend metadata. Idempotent: re-running keeps user-authored files and merges backend config instead of overwriting. No API keys required.",
@@ -462882,7 +462882,7 @@ var sceneRenderSchema = z.object({
   workers: z.number().optional().describe("Capture worker count (1-16). Default 1.")
 });
 var sceneRenderTool = defineTool({
-  name: "scene_render",
+  name: "render",
   category: "scene",
   cost: "free",
   description: "Render a scene project to MP4/WebM/MOV via the Hyperframes producer. Requires Chrome installed locally. Output defaults to renders/<projectName>-<isoStamp>.<format>.",
@@ -462899,7 +462899,7 @@ var sceneRenderTool = defineTool({
       workers: args.workers
     });
     if (!result.success) {
-      return { success: false, error: result.error ?? "scene_render failed" };
+      return { success: false, error: result.error ?? "render failed" };
     }
     return {
       success: true,
@@ -462937,10 +462937,10 @@ var sceneBuildSchema = z.object({
   force: z.boolean().optional().describe("Re-dispatch primitives even when cached assets exist.")
 });
 var sceneBuildTool = defineTool({
-  name: "scene_build",
+  name: "build",
   category: "scene",
   cost: "high",
-  description: "v0.60 one-shot orchestrator: read STORYBOARD.md per-beat YAML cues (narration / backdrop / duration), dispatch TTS + image generation per beat, compose scene HTML via the compose-scenes-with-skills pipeline, then render to MP4. Use this instead of chaining scene_init + scene_add + scene_render manually. Caches by SHA256 of (DESIGN.md + cue body) so re-runs are idempotent and cheap.",
+  description: "v0.60 one-shot orchestrator: read STORYBOARD.md per-beat YAML cues (narration / backdrop / duration), dispatch TTS + image generation per beat, compose scene HTML via the compose-scenes-with-skills pipeline, then render to MP4. Use this instead of chaining init + scene_add + render manually. Caches by SHA256 of (DESIGN.md + cue body) so re-runs are idempotent and cheap.",
   schema: sceneBuildSchema,
   async execute(args, ctx) {
     const projectDir = args.projectDir ? resolve23(ctx.workingDirectory, args.projectDir) : ctx.workingDirectory;
@@ -462960,7 +462960,7 @@ var sceneBuildTool = defineTool({
       force: args.force
     });
     if (!result.success) {
-      return { success: false, error: result.error ?? "scene_build failed" };
+      return { success: false, error: result.error ?? "build failed" };
     }
     return {
       success: true,
@@ -463043,7 +463043,7 @@ var sceneComposePromptsTool = defineTool({
   name: "scene_compose_prompts",
   category: "scene",
   cost: "free",
-  description: "Emit the per-beat compose plan for the host agent to author scene HTML itself. Reads STORYBOARD.md + DESIGN.md and returns each beat's outputPath + userPrompt + cues + body, plus references to the project's SKILL.md (Hyperframes rules) and DESIGN.md (visual identity). The host agent writes each compositions/scene-<id>.html file directly \u2014 VibeFrame makes NO LLM call here. Pairs with scene_install_skill (Phase H1). Phase H2 of the agentic-native composer plan; the internal-LLM batch path (scene_build) remains as a fallback for non-agent contexts.",
+  description: "Emit the per-beat compose plan for the host agent to author scene HTML itself. Reads STORYBOARD.md + DESIGN.md and returns each beat's outputPath + userPrompt + cues + body, plus references to the project's SKILL.md (Hyperframes rules) and DESIGN.md (visual identity). The host agent writes each compositions/scene-<id>.html file directly \u2014 VibeFrame makes NO LLM call here. Pairs with scene_install_skill (Phase H1). Phase H2 of the agentic-native composer plan; the internal-LLM batch path (build) remains as a fallback for non-agent contexts.",
   schema: sceneComposePromptsSchema,
   async execute(args, ctx) {
     const projectDir = resolve23(ctx.workingDirectory, args.projectDir);
@@ -463136,7 +463136,7 @@ var audioIsolateTool = defineTool({
   }
 });
 var audioVoiceCloneTool = defineTool({
-  name: "audio_voice_clone",
+  name: "audio_clone_voice",
   category: "audio",
   cost: "low",
   description: "Clone a voice from audio samples using ElevenLabs. Requires ELEVENLABS_API_KEY.",
@@ -463923,7 +463923,7 @@ async function executeSuggestEdit(options) {
 
 // ../cli/src/tools/manifest/analyze.ts
 var analyzeMediaTool = defineTool({
-  name: "analyze_media",
+  name: "inspect_media",
   category: "analyze",
   cost: "low",
   description: "Analyze media (image, video, or YouTube URL) using Gemini AI. Requires GOOGLE_API_KEY.",
@@ -463951,7 +463951,7 @@ var analyzeMediaTool = defineTool({
   }
 });
 var analyzeVideoTool = defineTool({
-  name: "analyze_video",
+  name: "inspect_video",
   category: "analyze",
   cost: "low",
   description: "Analyze video content using Gemini AI with temporal understanding. Requires GOOGLE_API_KEY.",
@@ -463979,7 +463979,7 @@ var analyzeVideoTool = defineTool({
   }
 });
 var analyzeReviewTool = defineTool({
-  name: "analyze_review",
+  name: "inspect_review",
   category: "analyze",
   cost: "low",
   description: "AI video review: analyzes quality, suggests fixes, and optionally auto-applies them. Requires GOOGLE_API_KEY.",
@@ -464007,7 +464007,7 @@ var analyzeReviewTool = defineTool({
   }
 });
 var analyzeSuggestTool = defineTool({
-  name: "analyze_suggest",
+  name: "inspect_suggest",
   category: "analyze",
   cost: "low",
   description: "Get natural-language edit suggestions for a project from Gemini. Returns suggestions array with type/confidence/clipIds. With `apply: true`, applies the first suggestion in place. Requires GOOGLE_API_KEY.",
@@ -465325,7 +465325,7 @@ async function executePipeline(manifest2, options = {}) {
 
 // ../cli/src/tools/manifest/pipeline.ts
 var pipelineHighlightsTool = defineTool({
-  name: "pipeline_highlights",
+  name: "remix_highlights",
   category: "pipeline",
   cost: "low",
   description: "Extract highlight clips from a longer video using AI analysis. Requires OPENAI_API_KEY+ANTHROPIC_API_KEY or GOOGLE_API_KEY (with --use-gemini).",
@@ -465358,10 +465358,10 @@ var pipelineHighlightsTool = defineTool({
   }
 });
 var pipelineAutoShortsTool = defineTool({
-  name: "pipeline_auto_shorts",
+  name: "remix_auto_shorts",
   category: "pipeline",
   cost: "medium",
-  description: "Automatically generate short-form content (Reels/TikTok/Shorts) from a longer video. Same API key requirements as pipeline_highlights.",
+  description: "Automatically generate short-form content (Reels/TikTok/Shorts) from a longer video. Same API key requirements as remix_highlights.",
   schema: z6.object({
     video: z6.string().describe("Path to the input video file"),
     outputDir: z6.string().optional().describe("Output directory for shorts"),
@@ -465395,7 +465395,7 @@ var pipelineAutoShortsTool = defineTool({
   }
 });
 var pipelineRunTool = defineTool({
-  name: "pipeline_run",
+  name: "run",
   category: "pipeline",
   cost: "very-high",
   surfaces: ["mcp"],
@@ -465474,7 +465474,7 @@ var pipelineRunTool = defineTool({
   }
 });
 var pipelineRegenerateSceneTool = defineTool({
-  name: "pipeline_regenerate_scene",
+  name: "remix_regenerate_scene",
   category: "pipeline",
   cost: "high",
   description: "Regenerate specific scenes against an existing storyboard.{yaml,json} on disk. Can regenerate video, image, or narration independently.",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vibeframe/mcp-server",
-  "version": "0.75.0",
+  "version": "0.76.0",
   "description": "VibeFrame MCP Server - AI-native video editing via Model Context Protocol",
   "type": "module",
   "bin": {
@@ -57,8 +57,8 @@
     "tsx": "^4.21.0",
     "typescript": "^5.3.3",
     "vitest": "^1.2.2",
-    "@vibeframe/cli": "0.75.0",
-    "@vibeframe/core": "0.75.0"
+    "@vibeframe/core": "0.76.0",
+    "@vibeframe/cli": "0.76.0"
   },
   "engines": {
     "node": ">=20"