CineForge-Agent MCP Server
Exports OpenTimelineIO (.otio) timelines that can be imported into DaVinci Resolve for further editing alongside the generated master MP4.
Uses FFmpeg to normalize master audio to -14 LUFS and produce final MP4 outputs.
Connects to local Ollama Qwen2.5-VL for VLM-based QA of rendered frames, detecting anatomy or prompt-fidelity issues and triggering prompt rewrites and retries.
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
CineForge-Agent
A Schema-First, Self-Healing Multi-Agent AI Video Production Studio with MCP 2.x, C2PA Provenance & NLE Export.
CineForge unifies agentic studio orchestration, long-form character/scene continuity, and deterministic HTML/GSAP motion graphics while closing key gaps in open-source AI video workflows:
Persistent
EntityBible: Locks character appearance tokens, seeds, environment lighting LUTs, and voice IDs across every shot and episode.VLM Self-Healing Critique Loop: Extracts frames from rendered clips, audits anatomy and prompt fidelity via local Ollama (
qwen2.5vl) or mock/fallback QA, and auto-rewrites failed prompts within a pre-render predictive$budget cap.Phoneme Lip-Sync & Hybrid Motion Graphics: Routes talking-head dialogue through MuseTalk while rendering kinetic typography and charts via headless Playwright + GSAP.
Native NLE Export (
.otio) & C2PA Provenance: Outputs-14 LUFSnormalized Master.mp4files alongside OpenTimelineIO (.otio) timelines for DaVinci Resolve and Premiere Pro. Generates an unsigned provenance sidecar (.unsigned_c2pa.json) by default, or embeds a cryptographically signed X.509 C2PA manifest whenC2PA_CERT_PATHandC2PA_KEY_PATHare configured.Built-in MCP 2.x Server: Drive the entire studio directly from Claude Desktop, Cursor, or VS Code Copilot (
MODE=mock,MODE=hybrid, orMODE=production).
System Architecture
flowchart TD
MCP["MCP 2.x Server (Claude / Cursor / VS Code)"] --> DAG["DAG Orchestrator (Checkpoints & Predictive Budget Caps)"]
BIBLE["EntityBible (Characters, Seeds, Voices)"] --> DAG
DAG --> TTS["Kokoro-82M TTS"] & T2V["ComfyUI (Wan 2.2)"] & HTML["Playwright HTML / GSAP"]
TTS & T2V --> LIP["MuseTalk Lip-Sync"]
LIP & HTML --> VLM["VLM QA Critic (Qwen2.5-VL / Runtime Fallback)"]
VLM -- "Failed (Rewrite + Retry)" --> DAG
VLM -- "Approved" --> C2PA["C2PA Provenance (Sidecar or X.509 Signed)"]
C2PA --> MP4["Master MP4 (-14 LUFS)"] & OTIO["OpenTimelineIO (.otio)"]Related MCP server: Gemini Omni MCP Server
Verification & Test Scope Matrix
Layer | Verification Method | Status |
Core Orchestrator, Schemas & NLE Export | Automated | Verified in CI |
Provider HTTP Protocols & Hybrid Runtime Fallback | Real TCP socket integration tests ( | Verified in CI |
Code Quality, Static Types & Containerization |
| Verified in CI |
Live GPU Model Inference (Wan 2.2, Kokoro-82M, Qwen2.5-VL, MuseTalk) | Requires local GPU/CUDA model weights & running services ( | External Runtime |
Quick Start
git clone https://github.com/Duttaak555/cineforge-agent.git
cd cineforge-agent
uv venv --python 3.12 && source .venv/bin/activate
uv pip install -e ".[dev]"
playwright install chromium
cineforge-doctor
pytest -v
python examples/run_demo.pyLicense
Apache 2.0 — Copyright 2026 Akashdeep Dutta (Duttaak555)
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
Plan, compare, price, generate, and recover AI video from compatible MCP clients.
Direct AI video, image, music and storyboard production from any MCP client.
AI video editor for agents and humans: timeline, captions, color, audio and generation as MCP tools.
Build, run, schedule, and publish AI video pipelines to YouTube and TikTok from any MCP client.
Related MCP Servers
- FlicenseNot gradedqualityDmaintenanceMCP server that provides a local Web production console and HTTP control plane for AI video creation, exposing tools for project/revision management, media analysis, editing, review, and delivery while keeping formal rendering behind explicit approval gates.-
- AlicenseNot gradedqualityDmaintenanceExposes video generation and editing tools to MCP-capable clients, enabling text-to-video, image-to-video, video editing, reusable voice and character profiles, and task status queries through natural language.4MIT
- AlicenseNot gradedqualityBmaintenanceEnables AI agents and MCP clients to run self-hosted video production workflows, including project planning, rendering, review, and delivery with durable background jobs and local review.2Apache 2.0
- AlicenseNot gradedqualityBmaintenanceEnables AI agents to direct a local media studio over MCP, generating consistent characters, photocards, album art, and beat-synced music videos by orchestrating ComfyUI, ffmpeg, and Piper TTS while recording the full recipe of every asset.MIT