AI assistants are part of Pro. Teleprompter Studio works with any assistant that supports MCP (the Model Context Protocol), an open standard for connecting assistants to apps.
What an assistant can do
- Find and read your scripts and folders.
- Rewrite a script for speaking aloud, add pause marks, bold, highlights and your delivery cues.
- Turn pause marks on for a script and set how long each one holds.
- Set a script’s pace (words per minute), when you ask it to.
- Work out how long a script takes to read.
- Create new scripts, with a title and folder.
- See your folders and create new ones (including a nest of folders, like “Courses / Module 1”) to put new scripts in.
It can’t delete, rename or move scripts or folders, see your recordings, change your settings, or touch your account.
You stay in control
Everything an assistant saves goes into one pending change on the script, shown in the app as “Changed by Claude · 2 edits” and marked AI edited in your library. If you use more than one assistant, they build on the same pending change. Then you choose:
- Keep: it becomes a saved version.
- Undo: the script goes back to your last saved version.
- Edit it yourself: your edit keeps it, with your changes on top.
Use Compare to see exactly what changed. Your last five saved versions are always in Versions, so you can restore any of them. A new script from an assistant shows as “Created by …” until you keep it.
Connecting an assistant
Add Teleprompter Studio to your assistant as a connector (sometimes called an MCP server or tool) with this address:
Server address
https://teleprompterstudio.affirmdigitalproducts.com/api/mcpThe assistant then opens a Teleprompter Studio page where you sign in and choose Allow. That’s it: no passwords are shared with the assistant. Step-by-step guides:
- Claude: claude.ai, the Claude desktop and mobile apps, and Claude Code.
- Gemini: Gemini CLI, and Gemini where Google supports connectors.
- ChatGPT: ChatGPT apps and connectors, and the OpenAI API.
- Ollama: Local models through an MCP-capable app, with an access token.
- Copilot: GitHub Copilot in VS Code, and Microsoft Copilot Studio.
- Siri: Siri and Shortcuts in the iPhone and iPad app, and Apple’s AI with a Shortcut.
Apps that can’t sign in themselves can use a personal access token instead: create one on My account under Connected assistants, and treat it like a password.
Try asking
- “List my Teleprompter Studio scripts.”
- “Tighten my Product launch script for speaking, add pause marks and highlight the key numbers.”
- “How long will Welcome video take at 140 words per minute?”
- “Write a 60-second intro for my course in the Courses folder.”
Many assistants also offer a ready-made prompt, Prepare for recording, that does a full pass on a script.
Disconnecting
My account → Connected assistants lists every assistant and token with when it was last used. Disconnect stops it working immediately. If Pro ends, assistants stop working and any unreviewed changes are undone.
Safety: text that talks to the AI
Words copied from a web page or document can hide instructions meant for an AI (“ignore your instructions and…”). This is a trick called prompt injection. Teleprompter Studio guards against it:
- Assistants are told that your scripts are content to work on, never instructions to follow.
- Invisible characters, a common way to hide such text from people, are removed from everything assistants read and save.
- Wording that looks like instructions to an AI is pointed out to the assistant, and if an assistant saves some, the change is marked ⚠ check this in the app so you can look at it in Compare before you keep it.
- Assistants can’t delete anything or send anything anywhere from Teleprompter Studio, and every change they make can be undone.
No filter catches everything. If you connect other tools to the same assistant (email, files, messaging), keep their “ask before acting” confirmations switched on, so a hidden instruction can’t quietly use them.
Privacy
When an assistant reads a script, the text goes to that assistant’s provider (for example Anthropic, Google, OpenAI or Microsoft) and is handled under their terms. A local model in Ollama keeps it on your computer. We never log your script text. See the Privacy Policy.