Takejam

Video Editing MCP Servers Compared (2026) | Takejam

Short answer: A video editing MCP server lets ChatGPT, Claude or another AI app edit video by calling an editor's tools. Pick by job: Runway mainly generates new clips; Kapwing, invideo and Vyra edit footage you upload to their cloud; Takejam edits a timeline while your files stay on your computer. Prices range from free to $24 a month and up.

What is a video editing MCP server?

MCP (Model Context Protocol) is the open standard that lets an AI app use other software. A video editing MCP server is an editor's "plug" for it. You paste one web address into ChatGPT, Claude or another MCP app, sign in, and the AI can now call that editor's tools: read the timeline, cut, add titles, render. If MCP is new to you, start with what MCP is, explained simply.

In the last two months this went from rare to crowded. invideo launched its MCP on 30 September 2026, Vyra rebuilt its server in October, CapCut announced a ChatGPT plugin on 24 October, and Kapwing and Runway already had theirs. They all say "edit video with AI", but they do very different jobs.

We build one of them (Takejam), so we'll be plain about our own limits too. Everything about other companies below comes from their own pages, checked on 17 November 2026.

Which video MCP servers are there, and how do they compare?

ServerMain jobAI apps namedWhere your footage goesFree option
TakejamEdit your own footage on a timelineAny MCP app: ChatGPT, Claude, Gemini and othersStays on your computerYes: 50 AI edits a day, 1080p, no watermark
KapwingEdit and generate (subtitles, dubbing, prompt to video, B-roll)Claude, ChatGPTUploaded to your Kapwing workspaceYes, but free exports are 720p with a watermark (pricing)
invideoTimeline editing plus image, video and audio generationClaude, ChatGPT, Codex, Cursor, VS Code, Gemini CLI and moreinvideo's browser workspaceManual editor is free; generation uses credits
VyraEdit your footage (doesn't generate footage)Claude, ChatGPT, Codex, CursorUploaded; edits run on Vyra's servers3-day trial, then from $24 a month
RunwayMainly generate new video and imagesChatGPT, Claude, Cursor and othersResults saved to your Runway libraryUses Runway credits
CapCut pluginBuilds a first cut from an idea, then hands off to CapCut's editorChatGPT onlyCapCutNot clearly published; see our CapCut post

Two things jump out. First, "edit" and "generate" are different products. Second, almost every option needs your footage in someone's cloud before the AI can touch it.

Do you need an editing MCP or a generating MCP?

Ask yourself one question: do you already have the footage?

  • You have clips (a talking-head video, an event, a product shoot): you need an editing server. The AI's job is cutting, ordering, captions, titles, colour and music, on footage you shot.
  • You have only an idea or a script: a generating server like Runway, or the generation side of Kapwing or invideo, makes new shots. Those use credits per generation, and a write-up on Runway's MCP warns about strict content moderation and long processing for video.

Most creators and teams we talk to are in the first group. Generated clips make great B-roll, but the main story is usually something real you filmed.

Where do your video files go?

This is the part comparison posts skip, and it matters for client work, unreleased products and long files.

With a cloud editor, the AI can only edit what has been uploaded. Kapwing says an upload "lands straight in your Kapwing workspace". Vyra's agent edits "on Vyra's servers". That's fine for many people, but it means uploading gigabytes, waiting, and keeping a copy of client footage on another company's servers. Kapwing's free plan also caps uploads at 2 GB and keeps projects for 3 days.

Takejam works the other way round. The editor runs in Chrome or Edge on your computer, and your media files stay there. What the AI reads through the connector is the timeline: which clip sits where, the titles, captions, effects and the music's beat grid. When the AI wants to see the result, its look_at_frames tool asks your open editor to render a few frames exactly as the export would, and only those frames go to the AI. More on this in editing video without uploading your files.

The trade-off: because the footage lives on your computer, the project has to be open in Takejam on someone's computer for the AI to look at frames or for you to export. A cloud editor can render with your laptop closed.

What can the AI actually do through the Takejam connector?

Here is the real tool list your AI sees after you connect https://api.takejam.com/mcp:

ToolWhat it does
list_projectsLists the projects in your workspace
get_projectReads the timeline: tracks, clips, media in the bin, scenes, markers, beat grid
edit_projectMakes edits as one undoable change: add, trim, split, move or delete clips, speed and speed ramps, keyframe animation, effects, transitions, templates, aspect ratio, caption styles, colour replacement, snapping cuts to the beat
get_referenceLooks up the available effects, looks, fonts, transitions, text animations and caption styles
look_at_framesRenders frames or a contact sheet so the AI checks its own work
get_review_notes / reply_to_noteReads the team's review notes and answers or resolves them
undo_last_ai_editPuts the project back to before the last AI edit

Each edit_project call is all-or-nothing: if one step fails, nothing changes and the AI gets the reason. Everyone with the project open sees the edit appear live, and Undo AI change in the top bar reverses it. Scenes your reviewers marked approved are locked, so the AI can't quietly change them.

What it can't do, honestly:

  • It can't upload or create footage. You drop media into the project; the AI edits what's there. It doesn't generate new video.
  • No phone editing. Takejam runs in Chrome or Edge on Windows, Mac and Linux.
  • The free plan has a daily cap of 50 AI edits. Creator ($5 a month) makes them unlimited.

How do you connect a video editing MCP server?

The steps are nearly the same for every server. Here they are with Takejam:

  1. Create a free account at app.takejam.com, make a project and drag your clips into the Media tab.
  2. In Claude: open Customize → Connectors, click + Add, then Add custom connector, and paste https://api.takejam.com/mcp. Claude's Free plan allows one custom connector (Anthropic's help page).
  3. In ChatGPT: open Settings → Apps & Connectors → Advanced, turn on Developer mode, click Create, paste the same address and choose OAuth.
  4. A Takejam sign-in page opens. Pick your workspace and click Allow.
  5. In a new chat, turn on Takejam and ask: "Cut my clips into a 30-second reel, add captions and a bold title at the start."

Full walkthroughs with every click and prompts that work: Claude, ChatGPT and Gemini. Gemini can also work without a connector: paste a free Gemini API key into Takejam's AI tab.

What does each one cost?

From each company's own page, 17 November 2026:

ServerFreePaid from
Takejam50 AI edits a day, 1080p, no watermark, 7-day historyCreator $5/month (unlimited AI edits, 4K); Team $8 per seat/month, minimum 2 seats
Kapwing10 credits, 720p, watermark, 4-minute exportsPro $16/month
invideoManual editor free; no free plan listed on the pricing pageBasic from $4 per seat/month, billed yearly
Vyra3-day trialMCP Starter $24/month (up to 1440p)
RunwayNot stated on the MCP pageCredits

Remember the AI app itself is billed separately by OpenAI, Anthropic or Google. And credit-based servers can run down faster than you expect: Kapwing's own MCP page says you "may deplete credits faster than expected" when an agent works.

Which one should you pick?

  • You want brand-new shots from a prompt: Runway, or the generation tools in Kapwing or invideo.
  • You already live in Kapwing or invideo: use their server; your projects are already there.
  • You want an AI to cut footage you shot, keep the files on your own computer, and still finish by hand on a real timeline: that's what Takejam was built for. It's free to try, and a team can watch the AI's edits land live in the collaborative editor.

Questions, answered

What is the best MCP server for video editing?

It depends on the job. For generating new clips, Runway. For editing footage already in a cloud workspace, Kapwing or invideo. For editing your own footage on a timeline while files stay on your computer, with any AI app, Takejam. Check whether you need edit or generate first.

Can ChatGPT edit videos with MCP?

Yes. In ChatGPT, turn on Developer mode under Settings → Apps & Connectors → Advanced and add a video editor's MCP address, such as https://api.takejam.com/mcp. ChatGPT can then cut, title and caption your project through that editor.

Is there a free video editing MCP server?

Yes. Takejam's connector works on the Free plan with 50 AI edits a day and 1080p exports without a watermark. Kapwing's MCP works on free accounts too, but free exports are 720p with a watermark.

Do I have to upload my videos to use a video MCP?

With most of them, yes: Kapwing, Vyra and other cloud editors need your files on their servers first. Takejam doesn't; your media stays on your computer and the AI edits the timeline.

Does Claude's free plan support video editing connectors?

Yes. Anthropic's help page says custom connectors work on every Claude plan, including Free, which is limited to one custom connector.

Can the AI see my video, or does it edit blind?

Good servers let it check. Takejam's look_at_frames tool renders frames exactly as the export would and shows them to the AI, so it can spot a title covering a face or a cut on the wrong shot and fix it.

Mahfuj Khondoker
Founder of Takejam · designer and video editor. Building Takejam, the video editor you can drive with any AI.