kinocut

---

What is Kinocut?

Kinocut is a free, open-source Model Context Protocol (MCP) 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.

Also known as	kino (CLI); formerly mcp-video / mcp_video
Latest published release	1.9.0 (2026-07-15)
Product site	kinocut.dev
PyPI	kinocut
MCP Registry	io.github.KyaniteLabs/kinocut
Canonical source	Forgejo — git.kyanitelabs.tech (source of truth; accepted commits land here first)
Public collaboration	GitHub — KyaniteLabs/kinocut (clone, stars, issues, discussions, security reports, CI)
License	Apache-2.0
Runs on	Your machine (macOS, Linux, Windows) — FFmpeg required on PATH
Not	A hosted cloud editor, credit-metered SaaS, or untyped FFmpeg shell wrapper

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 AI crawlers: llms.txt · site: kinocut.dev/llms.txt.

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

Surface	Version / tip	What it means
PyPI / npm / GitHub Release	1.9.0 (2026-07-15)	Latest published package. Install with pip install -U kinocut.
This repository (master)	151 MCP tools / 130 CLI commands	Development tip ahead of published 1.9.0 (release-artifact policy surfaces: review, publish gate, learning, cost, recipe, capabilities, benchmark).
Next	trusted execution kernel + sound program depth	See Whats next. Not pinned to a specific package version yet.

Install from PyPI for the stable package. Clone master only when you intentionally need post-tag tip work.

What's in 1.9.0 (latest release)

Kinocut 1.9.0 is what you get from pip install kinocut today (150 MCP tools / 129 CLI commands). 1.9.0 adds public MCP/CLI/client surfaces for eight release-artifact policy engines (review package, publish gate, review decisions, learning report, cost ledger, recipe capture, capabilities, benchmark) on top of the 1.8.0 contract-first AI-video foundation:

• Everything from the 1.7.0 identity cutover (kinocut package, kino CLI, MCP Registry id, kinocut.dev)
• Governed AI-video — content-addressed ingest, unified preflight, temporal inspect, exact-asset verdict / acceptance, body-swap, lineage-bound salvage (docs/AI_VIDEO_REVIEW_AND_SALVAGE.md)
• Optional C2PA on path-based export (off by default; verify-after-sign)
• Staged MCPB package foundations (Desktop install experiments; not a fully self-contained published runtime)
• Field safety — loss-proof add-audio policies; authored ASS + dimension-aware subtitles
• Hyperframes under MCP — hyperframes_init no longer hangs without a TTY
• Compatibility: mcp-video==1.6.2 installs kinocut==1.8.0; mcp_video imports, MCP_VIDEO_*, ~/.mcp-video, mcp-video://, and legacy receipt keys remain supported on the 1.8.x line

Also still on the published line from earlier releases:

• Agent workflow engine, video rescue, post-rescue planning, layered compositing, expanded preflight guardrails

Full notes: CHANGELOG.md · v1.8.0 release

Whats next

Post-1.8 product work (not a published package version):

Already in 1.8.0

Area	What landed on master	Start here
Governed AI-video	Content-addressed video_ingest, unified video_preflight, temporal evidence (video_inspect_temporal), exact-asset video_verdict / video_acceptance_eval, audio-preserving video_body_swap, lineage-bound video_salvage	docs/AI_VIDEO_REVIEW_AND_SALVAGE.md
Project store / contracts	Append-only private project storage, strict canonical records, protected-element checks, fail-soft optional visual providers	docs/AI_VIDEO_CONTRACTS.md · docs/AI_VIDEO_INSPECTION.md
Field safety	Loss-proof add-audio duration policies; authored ASS + dimension-aware SRT/VTT subtitles	CHANGELOG.md 1.8.0
C2PA provenance	Optional signing on path-based export / Client.export() via c2patool (off by default; only reports signed after verify)	docs/C2PA_PROVENANCE.md
MCPB packaging	Staged Desktop package + fail-closed native builder foundation; not a published self-contained runtime yet	docs/MCPB.md
Repurpose skill	Path-based skills/kinocut-repurpose + deterministic current-tools demo (marketing seed, not the final kernel-backed product)	docs/REPURPOSE_SKILL.md
Hyperframes under MCP	hyperframes_init no longer hangs without a TTY (non-interactive init + closed stdin)	CHANGELOG.md 1.8.0

Upcoming pipeline (in progress)

Two coordinated programs remain after the published 1.8.0 release:

1. AI-video + review/salvage finish

Contract-first media identity → inspection → human-gated verdict → bounded derivatives. Remaining work includes independent Wave-3 verification freeze, audio continuity, subtitle/graphics QA depth, asset intelligence, editorial planning, learning reports, and whole-program acceptance. Sequencing: wishlist parallel execution · current status: post-1.8 program status.

2. kinocut_sound (Sonic World) — full-episode audio production

Standalone-capable sound package inside this repo (kinocut_sound/): plan/timeline/routing/consent → voice → post/spatial → ambience/world → mix/stems → QA/metadata → thin public adapters → host joins → dual-class benchmark → STOP.

Slice	Focus	Status (as of 2026-07-14)
S1–S4	Contracts, authorization, registry/policy, script/episode planning	Implemented foundation leaves
S5 / S7 / S8	Base voice, post/spatial chain, ambience/world	Integrated leaves on master
S6 / S10	Consent-gated clone/blend; voice consistency	Integrated leaves on master
S9 / S11	Mix assembly/stems; QA + metadata	Integrated leaves on master
S12	Thin public discovery / Python adapters	Integrated (capability discovery surface)
S13	Kinocut/host joins (D41/D42 production bindings)	Blocked — external owner receipts incomplete
S14	Dual-class benchmark (Apple silicon + x86 Linux)	Partial — x86 available; Apple class external_host_unavailable
S15	Adversarial acceptance + release STOP	STOP — no ship without dual-class S14, S13 receipts, independent review, human authorization

Authoritative receipts: sound program handoff · S13–S15 gate · sound plan index.

3. Trusted execution kernel (post-program, gated)

The approved trusted execution layer plan still defines the durable product path after the current program: durable edit projects, async render/resume wrapping video_workflow_*, receipt lin