Kinocut

kinocut.dev

Free plan
Windows
macOS
Linux
API

What Kinocut is

Kinocut is a local-first MCP server, Python library, and command-line tool for AI-agent video editing. Its workflow turns local interviews into captioned vertical clips and can produce a JSON Video Receipt recording the request, tools used, edits, guardrails, and items needing human review. The published version provides 203 structured MCP tools for trimming, merging, subtitling, reframing, and audio normalization. Server-side validation checks risky or invalid parameters before FFmpeg renders. Quality gates can assess VMAF, loudness, black frames, and audio/video sync, holding output below a configured bar. AI highlight detection, automatic reframing, and social platform presets are supported. The maker says FFmpeg and Whisper run locally, media stays on the user’s machine, and the core needs no account or API key or telemetry. Integrations include Claude Code, Cursor, generic stdio MCP clients, Python automation, and CI. Kinocut runs on macOS, Ubuntu/Debian, and Windows. FFmpeg must be installed separately.

Who it suits

Kinocut suits AI agents, automation workflows, and agency teams needing local editing with review before client delivery. It is also intended for users connecting MCP clients or Python and CI automation.

Good to know, and what to watch

203 structured MCP editing tools
Local FFmpeg and Whisper processing
JSON receipts document workflow changes
Quality gates check video and audio
No account or API key required

FFmpeg must be installed separately
No hosted multi-tenant SaaS editor
No SOC 2 or ISO certificate package
Human review remains required for publishing

Plans

Every plan and price

Free core

Free

Free plan

The verdict

Kinocut combines structured AI-agent editing tools with local processing, validation, receipts, and quality gates. It fits privacy-conscious automation that keeps a review step before delivery.

What the maker publishes

Core workflow: It is designed to turn local interviews into captioned vertical clips with a Video Receipt. source
Dependencies: FFmpeg must be installed separately and available on PATH; optional Hyperframes support requires Node.js 22 or later. source
Editing tools: The published version provides 203 structured MCP tools for operations including trimming, merging, subtitling, reframing, and audio normalization. source
Integrations: The maker documents integrations with Claude Code, Cursor, generic stdio MCP clients, Python automation, and CI. source
Intended users: The maker describes it for AI agents, automation, and agency teams that need local video editing with review before client delivery. source
Limitations: The maker says it does not provide a hosted multi-tenant SaaS editor, a SOC 2 or ISO certificate package, or fully automatic brand-safe publishing without human review. source
Local processing: The maker says media stays on the user's machine and FFmpeg and Whisper run locally. source
Platform support: The install page gives setup instructions for macOS, Ubuntu/Debian, and Windows. source
Preflight: Server-side parameter validation catches risky or invalid editing inputs before FFmpeg renders. source
Privacy claims: The maker states that the core requires no account or API key and has no analytics or phone-home telemetry. source
Purpose: Kinocut is a local-first MCP server, Python library, and CLI for AI-agent video editing. source
Quality checks: Quality gates can check VMAF, loudness, black frames, and audio/video sync, and hold outputs below the configured bar. source
Receipts: Each workflow can produce a JSON Video Receipt recording the request, tools run, changes, guardrails, and items needing human review. source
Support: The maker identifies community support through Forgejo or GitHub issues and says no paid SLA is claimed. source

Checked against the maker's pages on 7 Oct 2026.