Edits for macOS opens more than 300 editing tools to agents via MCP

Assembling a short clip from several videos, removing pauses, fixing the order of shots: these are tasks you want to hand to an agent without a long explanation of the interface. The author of the Telegram channel Tips AI described the macOS version of Edits, where, according to his description, this can be done through a built-in MCP. The agent gets the editor's tools, while the human sees the changes on the timeline. For a small team, a very concrete question arises: does the time from raw footage to a finished clip shrink?

What connecting an agent provides

The original post mentions more than 300 tools available via MCP, and ready-made commands for connecting Claude, Codex or Muse in the app settings. The author describes the editor itself as a free alternative to CapCut with no watermarks and no export limits. This information comes from a user review; availability and terms should be checked on the Edits website.

MCP is needed here as a way to pass commands from the agent to the application. The user states the task in words, the agent chooses the available operations, and the editor applies the changes. The mere existence of such a connection does not show how well the model understands an editing task or how well it selects shots.

The practical value of a visible timeline is the ability to check intermediate results. If the agent placed the wrong fragment or cut a line too abruptly, a person can notice it before the final export. For a team where one person prepares a draft and another approves the work, such a check may prove convenient. This is a possible use case, not the result of a published test.

How to test the value on a single clip

Imagine a task: a founder has a recording of a product demo and several separate shots of the interface. They need a short video for a launch page. An example instruction for the agent: "Keep the explanation of the main feature, remove long pauses, place the interface shot next to the corresponding line." This illustrates a working request, not a promise that Edits will carry it out without errors.

For a first check, one familiar project is enough. The desired result, the usual assembly time and the typical difficulties should already be clear. Then it is easier to see whether the agent helped or only shifted the work from manual editing to correcting its own decisions.

It is worth timing the whole process: preparing the instruction, waiting, reviewing, fixing and exporting. It is also useful to record separately how many times the agent chose the wrong fragment, broke the order or needed an additional explanation. A quickly assembled draft helps little if you then have to spend a long time restoring the clip's meaning.

For a developer, such a trial will show how predictably the agent uses the tools. For a product manager, whether a colleague can repeat the process without constant help. For a founder, how much of their own time was actually freed up. The number of available operations does not by itself answer any of these questions.

Where the review's conclusions end

The author of Tips AI writes that in his experience, editing through Edits was roughly ten times slower and worse than the agent working through ffmpeg. The post has no set of source files, identical tasks, time measurements or quality criteria. So this is an impression from a few attempts, not a comparative test on which a production workflow can be chosen.

The limitations on complex animation and the comparisons with other editors should also be read as the author's assessment. Without checking specific features, one cannot conclude which tasks the app fundamentally does not support. The claim about a paid license for MCP in DaVinci Resolve also requires clarification of the specific integration.

There is a reason to try Edits: controlling the editor through an agent can simplify preparing drafts that are convenient to review and edit right away. The decision to adopt it is best made after one complete working task. If, including review and fixes, the clip is ready faster, the tool is useful. If time goes into explaining every action, it is worth rethinking the task or the automation method.