- 1credit per minute of video
- 0charged when no clips come back
- 3videos processing at once
Works on any plan with credits
No Pro-only gate. If you have credits, your agent can use them, at the same credit cost as the web app.
MCP server · Claude, ChatGPT, Codex
Your AI agent, now a video editor.
Paste a podcast link into Claude, ChatGPT, Codex or Cursor and get ranked, captioned short clips back without leaving the chat.
- 1
In Claude, open Customize, then Connectors, then + Add, then Add custom connector.
- 2
Name it FrameOS and paste the server URL.
https://frameos.studio/mcp - 3
Keep "Use Claude's published identity", click Add, then Connect and sign in to FrameOS.
- 4
Ask: "What is my FrameOS credit balance?"
Works on the web, in Claude Desktop and on mobile once added. Free Claude plans allow one custom connector.
No API keys: you sign in to FrameOS once and your agent works inside your account.
Ask in the chat. Watch it happen.
One conversation: a long episode in, ranked clips out, captions restyled, and a post that waits for your yes.
Claude FrameOS
Waiting for a video linkBuilt for creators, not just developers.
Nothing posts without your yes
Drafting never posts. Every post shows the account, text and visibility first.
Starts the slow part first
The render begins right away, and your agent sorts out captions and platforms while it runs.
Five frame shapes
Say the platform and your agent picks the shape.
One sign-in, no API key
Sign in to FrameOS once, in your browser. Your agent works inside your account.
Everything you'd ask an editor.
Your agent does each one with the FrameOS tools. You just talk to it.
Turn this episode into 5 TikToks: [link]
Start with one prompt.
Works on podcasts, interviews, webinars, streams and sales calls. Click one to copy it.
Clip
Turn this podcast into 5 vertical shorts for TikTok: https://youtube.com/watch?v=...
Feed posts
Make 3 square clips for LinkedIn from this interview, focused on what we say about hiring.
Captions
Use the Karaoke style on all my clips from today, move the captions up a bit, and export them.
Quotes
Find the three most quotable lines in last week's episode, with timestamps.
Thumbnails
Make a 16:9 YouTube thumbnail for my top clip in the style of this one: [image link]
Content pack
Turn this webinar into a week of content: 7 clips, captions, thumbnails and posts for YouTube and LinkedIn.
· — FAQ
Good to know
What is the FrameOS MCP server?
It connects your AI agent to your FrameOS account. Add it to Claude, ChatGPT, Claude Code or Codex by URL and your agent can send a video for clipping, wait out the render, restyle captions, search the transcript, make thumbnails and draft posts, all inside your account.
Which AI apps does it work with?
Claude (web, desktop and mobile), ChatGPT, Claude Code, Codex, Cursor and VS Code with GitHub Copilot are tested end to end. Gemini CLI and Perplexity are in beta: sign-in is set up for them and their steps are on this page, but we're still testing each one end to end. If one gives you trouble, tell us at support@frameos.studio.
Which FrameOS plans does it work on?
Any plan with credits. There's no Pro-only gate, and your agent spends the same credits you'd spend in the web app: 1 credit per started minute of source video, charged only when clips are delivered, and 10 credits per thumbnail. Exports, caption changes, copy drafts and posting are free. Clips made on the free plan carry the FrameOS watermark, just like in the web app, and sending a link that already finished starts a new project that is charged again in full. Plans are on the pricing page.
Do I need an API key?
No. You sign in to FrameOS once, in a normal browser window, and your agent works inside your account. There is no key to copy or leak.
Can my agent post without asking me?
Drafting copy never posts. Posting is its own FrameOS tool, marked as one that changes things, so Claude, ChatGPT and Claude Code ask for your OK before they run it, unless you have told them not to ask. Posts go only to accounts you connected in FrameOS yourself.
How long does a render take?
Usually 10 to 30 minutes, depending on the video's length and how many clips you ask for. Agents that can wait check back on their own; chat apps tell you when to ask again. Up to 3 videos can be processing at once in a workspace.
What can't it do yet?
Scheduling posts, share links, dubbing, censoring, free-form timeline edits, choosing the clip length, and posting to TikTok or X (your agent can still write the copy). Those live in the FrameOS web app or are coming later.
Give your agent a video editor.
Add FrameOS to your AI app, sign in once, and ask for your first clips.