Point at any video and the full teardown comes back where you already work — the hook, the pacing, the shot list, the why. One video studied deeply, or the pattern across your whole reference library. Setup takes about sixty seconds.
Two ways teams use it · 01
You're shipping a launch video and there's one reference that just works. Paste the link mid-conversation — the breakdown comes back as data your tools can reason over, not a PDF they can only summarize.
Then keep going: › draft a 30-second script that follows the same structure — the blueprint, transcript and voice profile are already in the conversation.
Two ways teams use it · 02
You run teardowns every week and the value is in the throughline. Ask across everything your team has analysed — openings, pacing, structure — instead of one tab at a time.
One more step bundles them: › get_context_bundle · hooks_only — the openings and scorecards of several analyses, merged into a single answer-ready payload.
Why it reads differently · the part nobody else has
Most tools hand over metadata. videngineer returns the raw evidence and the read on what makes the video work — from one link, in one payload.
raw evidence on the left · the strategist's read on the right · both come back through the connector
On your terms
Setup · about sixty seconds
Runs locally next to your tools. Three steps — a key, the connector file, one config entry.
In your videngineer account: Account → API access → Generate key. Shown once; rotate or revoke it anytime from the same card. No account yet? Start free — any signed-in account can issue a key.
One small file plus its runtime. Download videngineer_mcp.py, then:
pip install "mcp[cli]" # or: uv add "mcp[cli]" curl -O https://videngineer.com/mcp/videngineer_mcp.py
Paste this into your MCP config — Claude Desktop (claude_desktop_config.json), Claude Code (.mcp.json in your project) or Cursor (Settings → MCP) — with your path and key:
{
"mcpServers": {
"videngineer": {
"command": "python",
"args": ["/absolute/path/to/videngineer_mcp.py"],
"env": { "VIDENGINEER_API_KEY": "your-key" }
}
}
}
Restart the app and five tools appear: analyze_video · get_status · get_report · list_analyses · get_context_bundle. Start with a link you wish your team had studied sooner.
Prefer zero setup?
Paste a URL in your account settings and you're connected — no file, no terminal. Early access also covers recreate-a-cut as it rolls out. One email when your spot opens.
Coming next
Two things in the works — honestly early, and the early-access list hears first.
Questions, answered straight
Claude Desktop, Claude Code and Cursor are the three we test against — plus anything else that speaks MCP, the open standard this connector is built on.
The connector is free. It runs on your videngineer account, so analyses use your plan's credits exactly like the app does — no separate API pricing to think about.
Yes. The connector is read-only against your own account and every id is checked server-side against your key. An analysis you don't own comes back as a redacted preview — title and thumbnail only.
An open standard that lets the apps above use outside tools directly. You paste one config entry and videngineer becomes something they can call — no code, no API to learn.
Stuck on setup? Write to [email protected] with the subject “MCP connector” and we'll get you running.