AI-Assisted Editing And MCP
Use Fraime.it with local agents to inspect projects, propose edits, and export.
Fraime.it exposes structured timeline operations for AI-assisted editing. The goal is to let an agent work with project data while the macOS app stays in control of the project, preview, undo, save, and export path.
Use AI-assisted editing when you can describe the outcome you want more easily than you can manually perform every timeline operation.
What agents can do
Agents can inspect clips, tracks, annotations, subtitles, overlays, visual assets, and export settings. They can create working copies, add titles or callouts, adjust timeline elements by ID, suggest trims, and help prepare exports.
Assistant edits are review-first.
Use the panel to inspect the current time and selection, ask for common edits, generate smart suggestions, and review proposed changes before applying them.
Good requests include:
- "Create a 30 second launch clip from this project."
- "Add a title card, trim the intro, and suggest where callouts would help."
- "Find filler words and typing-heavy sections I can speed up."
- "Prepare a vertical Short, but show me the edit plan before applying it."
Working copies
Agent-created projects use a visible .agent naming convention so the original recording stays separate. Use a working copy when asking for broader edits such as a launch demo, short product clip, or cleaned-up walkthrough.
Use the original project for archival source material. Use working copies for experiments, aggressive trims, or alternate exports.
Preview before apply
Agents can dry-run edits and report what would change before applying the plan. Review the plan, then apply only the edits you want.
Read the summary before accepting a plan. Pay special attention to edits that affect timing, export settings, redactions, subtitles, or large ranges of the timeline.
Smart edit suggestions
Fraime.it can suggest filler-word cuts from transcript words and speed-ups for silent typing stretches. These are reviewable suggestions, not automatic destructive edits.
Suggestions should still be judged like human edits. A filler word may be part of natural speech, and a silent typing section may still be useful if the viewer needs to see the code appear step by step.
Local control
The local bridge is designed so Fraime.it owns the source project and export. The agent proposes structured edits, and the app applies them through the same timeline model used by the editor.
What not to delegate
Do not rely on an agent as the final privacy reviewer. Always manually check redactions and exported videos before publishing. Also avoid vague prompts like "make it better" when you need a specific outcome; describe the target audience, length, format, and tone.
Troubleshooting
If an agent cannot find the project, make sure the project is open in Fraime.it and the local bridge is running. If an edit plan is too broad, ask for a smaller plan first, such as only title cards or only filler-word suggestions. If the result is not what you wanted, return to the original project or a previous working copy.