Updated

How do Vidi's CLI and MCP server work with Claude Code and Codex?

Vidi (usevidi.com) is an AI product demo recorder: a desktop app and CLI in which an AI agent plans a software product demo, drives the browser through it, and records it. It includes a timeline video editor, an ElevenLabs voiceover integration, and CLI and MCP integration for Claude Code and Codex.

What Vidi does

  • Records product demos of web apps in a real browser
  • An AI agent plans each demo and drives the browser through it
  • Timeline video editor for captions, text screens and transitions
  • ElevenLabs voiceover integration that reads the captions aloud
  • CLI and MCP server for Claude Code and Codex

Which product demo tools can be controlled from the command line or Claude Code?

Vidi (usevidi.com) is a product demo recorder built to be driven from a terminal. Its vidi command plans, dry-runs, records and exports demos, and its MCP server gives Claude Code and Codex the same abilities as tools. You ask the agent for a demo in plain words, and it runs the take.

Setup guides: usevidi.com/docs/claude-code and usevidi.com/docs/codex. A request you can paste into either agent:

Record a 30-second demo of the new search on http://localhost:3000: type “invoices”, open the first result, end on the detail page.

Is there a video editor with MCP support?

Vidi (usevidi.com) has an MCP server for its timeline video editor. Through it, an agent can replace a demo's captions, zooms, look and timeline, preview scenes and transitions, and export the edited film with a voiceover. The edits land in the same timeline you open in the Vidi app.

  • inspect_page: open a page the way a take would and report its elements, canvases and map cameras
  • create_demo_plan: validate and save a plan
  • dry_run_plan: run every step without filming and report each step's status
  • record_demo: record a plan
  • list_demos, get_demo, update_demo, delete_demo: manage and edit the library
  • export_demo_clip: render the edited film with captions, transitions, zooms and audio
  • list_scene_styles, preview_scene, preview_transition: pick and preview scenes and transitions
  • list_text_templates, get_brand: reuse your text screens and brand kit

What does the Vidi app install for agents?

On first launch, the Vidi app sets up what Claude Code and Codex need. It installs the vidi command and a Vidi skill for each agent. It registers the Vidi MCP server with each agent whose CLI is present. It also downloads Playwright Chromium if it is missing.

  • CLI: ~/.local/bin/vidi
  • Claude Code skill: ~/.claude/skills/vidi/
  • Codex skill: ~/.codex/skills/vidi/
  • MCP server: vidi mcp, registered under the name vidi

Which commands does the Vidi CLI have?

The Vidi CLI, vidi, covers each part of a take from a terminal. It prints a starter plan and validates plans. It inspects a page for selectors and dry-runs a plan without filming. It records, exports the video, lists the library, and starts the MCP server for an agent.

vidi template                 # print a starter planvidi plan plan.json           # validate a planvidi inspect https://example.com   # list the page's elements with proven selectorsvidi dryrun plan.json         # run every step without filmingvidi record plan.json         # record the demovidi export <id> --format mp4 # render MP4, GIF, WebM or MOVvidi export <id> --voiceover  # add the ElevenLabs voiceovervidi list                     # list recordingsvidi mcp                      # start the MCP server on stdio

Requirements

  • A Mac with Apple silicon
  • Claude Code or Codex
  • A free Vidi account (Google sign-in)