CutAgent launches for Resolve
Jindra Tilkovský launched CutAgent, an AI agent powered by Claude that automates repetitive tasks inside DaVinci Resolve and claims to save dozens of hours per project for pro editors. The product targets high-volume editing chores inside Resolve, hinting at a new class of assistant agents embedded in NLEs. (x.com)
CutAgent is published as an “agent‑first” video cutting library that represents edits as declarative JSON edit lists and performs media operations through FFmpeg. (github.com) Every CLI command in the CutAgent project outputs structured JSON designed as an Edit Decision List (EDL) format so operations are composable and machine‑readable for downstream tools. (github.com) Bridging that output into DaVinci Resolve is already practical: multiple Model Context Protocol (MCP) servers expose Resolve’s scripting API for timeline edits, markers, color operations and Fusion composition control, and a pip package for a Resolve MCP integration lists compatibility with Resolve 18+ in its install notes. (github.com) Anthropic’s Claude platform supplies an Agent SDK and a Task tool for orchestrating multi‑step agent workflows and running code — capabilities that let a Claude‑powered agent generate EDLs, call FFmpeg, and then drive an MCP bridge into Resolve. (platform.claude.com) Related open‑source agent skills (for example, a “cut‑motionless‑video” skill that detects static segments and trims via FFmpeg) illustrate reusable primitives — motion/silence detection, JSON EDL output, and programmatic handoffs — that CutAgent’s JSON approach is designed to interoperate with when connected to Resolve via MCP. (github.com)
