COEY launches AI-first video editor GroundSlate

COEY has launched GroundSlate, a free Mac video editor built for AI assistants instead of manual timelines. The app exposes 385 local MCP tools for editing, generation and rendering on Apple Silicon, aiming to move more of the video workflow onto users' own machines.

Why it matters: - GroundSlate shifts video editing toward an AI-agent workflow, which COEY says can reduce manual clip-cutting and put more of the creative process on local Macs. - The app is free at launch and runs on the user’s own machine, with no generation credits, which could lower the cost of AI-assisted video production. - The product is designed to let assistants search, assemble, generate and render a finished film from a user’s plain-language prompt.

What happened: - COEY announced GroundSlate™, an AI-first video editor built around Model Context Protocol, on September 9, 2026. - The app exposes 385 local tools to assistants including OpenAI’s ChatGPT, Anthropic’s Claude, Cursor and Google’s Gemini. - GroundSlate is available today at no cost for Mac users with supported Apple Silicon hardware.

The details: - GroundSlate is built as an MCP server with a media-centric interface that shows users what the assistant is seeing and doing. - The app has no traditional timeline, unlike editors such as Adobe Premiere Pro, Final Cut Pro, DaVinci Resolve and CapCut. - A preferred AI assistant can read the library, choose moments, assemble sequences, add visual effects, generate missing footage and render the final file. - GroundSlate is built exclusively for Apple Silicon and uses the MLX framework. - COEY says the app is optimized for the latest Mac mini and Mac Studio models with larger shared memory. - The app works on any M-Series Mac with at least 16GB of shared memory. - GroundSlate can generate video, music, sound effects and speech locally using open source models. - Users can clone a voice from a short sample or create a new one, then synthesize narration in that voice. - The app can generate shots that were never filmed, score sequences with music that automatically ducks under narration and drive a character’s mouth and expression to a recorded take. - Each generated asset includes a provenance record listing the model, prompt and seed. - GroundSlate can also store assets generated from AI media cloud providers, letting users mix frontier models with local models in one app. - The app indexes transcripts with word-level timing, scene descriptions sampled as finely as every quarter second, visible people, beat and section maps for music, plus location and time metadata. - Assistants can search for what was said, who was on camera, where a clip was shot and when, then return exact time windows instead of whole clips. - GroundSlate stores file information in JSON and markdown in an open catalog that agents can understand and use.

Between the lines: - COEY is betting that the next stage of editing will be less about manual timeline work and more about natural-language direction and automated decision-making. - The local-first architecture also suggests a push toward privacy, speed and lower per-use costs compared with cloud-heavy generation. - Murphy’s comments frame frontier models and local models as complementary rather than competing layers in the same workflow.

What's next: - COEY says connecting GroundSlate to frontier models like ChatGPT Astra or Claude Fable can unlock more powerful edits and motion graphics through the MCP layer. - The company expects more of the AI pipeline to move local as open models improve. - More information is available at the company's website.