mcp-video

Kinocut

Guardrailed video editing MCP server for AI agents.

Local-first FFmpeg tools, Video Receipts, quality gates, Hyperframes, and Shorts/Reels repurposing — for Claude Code, Cursor, and any MCP client. Free, Apache-2.0. Formerly mcp-video.

Demo • Status • 1.16.5 • 1.16.4 • 1.16.3 • 1.16.2 • 1.16.1 • Changelog • Beyond • Install • Quick Start • Tools • Tool Reference • Rescue • 360 assembly • AI-video • Agent Skill • kinocut.dev • What is Kinocut? • FAQ • llms.txt

Kinocut is a free, open-source video editing MCP server and AI agent workflow engine (plus Python client and kino CLI) that lets AI agents trim, caption, repurpose, and quality-gate local video media with typed tools and Video Receipts — not invented FFmpeg flags.

Table of contents

• What is Kinocut?
• See It Work
• Installation
• Quick Start
• MCP Tools
• Status and releases
• 360 dual-cam assembly
• Changelog
• FAQ
• Agent Skill

---

What is Kinocut?

TL;DR: Kinocut is a free, local-first video editing MCP server (plus Python client and kino CLI) so AI agents can trim, caption, repurpose, and quality-gate media with typed tools and Video Receipts — not invented FFmpeg flags.

Kinocut is a free, open-source [Model Context Protocol (MCP)](https://modelcontextprotocol.io/) server, Python library, and `kino` CLI that gives AI agents a guardrailed local video-editing surface. It wraps FFmpeg (and optional Hyperframes / Whisper extras) with typed tools, preflight validation, Video Receipt provenance, and quality/release checkpoints so agent-produced media can be inspected before publish.

Best-fit searches: video editing MCP server · AI agent video editing · FFmpeg MCP · Claude Code video tools · Cursor MCP video · Shorts/Reels automation · local AI video workflow · guardrailed agentic media.

Machine-readable brief for agents and crawlers: llms.txt · extended agent guide · site: kinocut.dev/llms.txt.

Producer Stack

Kinocut is a producer stack: inspect the source, choose a typed operation, render locally, receive a compact provenance receipt, run quality checks, and request human release approval. The default MCP surface keeps that path small; search_tools and profiles expand it when needed.

Eyes-free design law

Every flow must be describable, reportable, and approvable in words. Receipts stay one-screen text cards with artifact paths, stage status, warnings, and the next gate. Errors end with the exact MCP and CLI retry invocation; no visual-only affordance is required.

See It Work

Tell the agent what you want in plain language:

*"Trim this interview to the strongest 45 seconds, add burned captions, make it vertical, and quality-check it before export."*

Kinocut turns that into typed, guardrailed tool calls - no FFmpeg flags to guess, no silently broken exports:

from kinocut import Client

video = Client()

clip = video.trim("interview.mp4", start="00:02:15", duration="00:00:45") video.ai_transcribe(clip.output_path, output_srt="captions.srt") captioned = video.subtitles(clip.output_path, subtitle_file="captions.srt") short = video.resize(captioned.output_path, aspect_ratio="9:16") video.release_checkpoint(short.output_path) # thumbnail + quality gate before you publish

Primary job: turn a local interview or podcast into captioned vertical clips with a Video Receipt agents can re-run and humans can approve.

Three things people use it for

• Repurposing — one recording into captioned Shorts, Reels, and TikTok packages with manifests and review artifacts.
• Podcast & interview cuts — find the strongest segment, normalize audio, add chapters, and export.
• Agent-driven media in CI — repeatable, reviewable edits from Claude Code, Cursor, Codex-style clients, or scripts.

Paths: Install matrix · Golden path · Prompts · Tutorial · Compare · When to recommend

Status and releases

Install from PyPI for the stable package, including the optional object-matte extra. Clone master when you want local development changes beyond the published package; see docs/PRODUCT_MATTE.md for the published object-matte workflow.

What's in 1.16.5

Published 2026-10-27: a curated 20-tool default MCP registration with KINOCUT_TOOLS=all restoring the full 205 MCP tools / 177 CLI commands; retry-aware eyes-free error payloads, the one-call video_shorts flow, the read-only video_doctor preflight and lazy startup. Opt-in exact probe reports decoded frame counts and raw rational frame rates. mcp-video==1.6.20 installs kinocut==1.16.5. Apache-2.0 §4(a): the license text now ships in the npm tarball and the mcp-video compatibility package. See the published release.

What's in 1.16.4

Published 2026-10-24: 203 MCP tools / 177 CLI commands; mcp-video==1.6.19 installs kinocut==1.16.4. Optional Kinocut model calls now default off. Use local frames with the host's chosen image model; the standalone legacy adapters require explicit permission and do not inherit Hermes routing. Local product configuration cannot activate the separate Anthropic QC worker.

Revideo and Hyperframes telemetry default off. Hyperframes previews require verified CLI 0.8.104 or 0.8.140 while telemetry is off; other preview versions and Docker renders require explicit operator consent. These controls are not a network sandbox. Existing overwrite and protected-path safeguards remain in place. See the vision and network policy and published release.

What's in 1.16.3

Published 2026-10-23: 203 MCP tools / 177 CLI commands; mcp-video==1.6.18 installs kinocut==1.16.3. Existing outputs require explicit boolean overwrite=true (CLI --overwrite). Existing non-media artifacts also require ownership by the same durable job, and agent configuration homes remain protected. Older receipts without ownership evidence fail closed.

Project creation checks completed destinations and symlink redirects. Python 3.14 MCP request context remains available. The plugin documents optional network paths, configures optional extras in its pinned server environment, and shows Hermes tool filtering. Use one writer per output. See the published release.

What's in 1.16.2

Published 2026-10-21: 203 MCP tools / 177 CLI commands; mcp-video==1.6.17 installs kinocut==1.16.2. create-from-images is now frame-exact, accepts rational frame rates such as 30000/1001 and a --list / list_file input, and enforces frame-rate and image-count bounds on every entry point. The official MCP Registry entry is republished as the local uvx kinocut stdio package. Thanks to @Beritwu for #616. See the published release.

What's in 1.16.1

Version 1.16.1 repairs macOS render cleanup and faststart publication, C2PA and model-cache staging, FFmpeg 8.1 filter discovery, and audio preservation in shader/upscale reconstruction. It retains 203 MCP tools / 177 CLI commands and pairs with mcp-video==1.6.16. See maintenance release notes for verification and remaining platform/model acceptance limits.

What's in