MCP server built in

Dope product videos, directed by your AI agent

Dope MCP is a desktop screen studio with an MCP server inside. Claude Code, Cursor, or any MCP agent clicks through your product, drives the camera, writes the script, and voices it with your own TTS keys, in every language you ship.

Download for macOS Download for Windows Read the docs

Free during early access · macOS & Windows · Works with any MCP client

Record a 30-second demo of the new checkout flow

If it speaks MCP, it can direct your videos.

Claude Code
Cursor
Windsurf
Cline
Zed
Continue
GitHub Copilot
Gemini CLI
Codex CLI
Roo Code
+ any MCP client

"dope-mcp": { "command": "dope-mcp" }

One line in your MCP config. No plugin, no SDK.

How it works

Prompt it. Watch it. Ship it.

Prompt your agent

Tell Claude Code or Cursor what you want: the flow to show, the tone, the length, the languages. That's your whole job.

The agent directs

Over MCP it drives your app and the recorder together: navigating, typing, clicking, zooming the camera, dropping captions right on beat. Bad take? It re-records the scene, not you.

Export everywhere

The agent writes the narration, your TTS voice reads it, and one recording renders with captions and voiceover in every language you asked for.

The MCP loop, live

Watch an agent shoot a demo.

A stylized replay of a real session: the agent's tool calls on the left, the recorder responding on the right.

agent · claude code

Record a 30-second demo of the new analytics page. Zoom into the revenue chart, add captions, and narrate it in English and Japanese.

record.start { res: "1080p" }
browser.goto { url: "/analytics" }
mouse.click { target: "Last 90 days" }
camera.zoom { on: "#revenue-chart", scale: 1.6 }
caption.show { text: "Filters update the chart live" }
voice.narrate { lang: "en" }
camera.pan { to: "#breakdown" }
export.render { langs: ["en", "ja"] }

app.acme.dev/analytics

Analytics

30 days Last 90 days Custom

Revenue

$48.2k

+12%

Active users

9,381

+6%

Churn

1.9%

−0.3%

Revenue

USD

Top sources

Organic41%

Paid33%

Referral26%

Breakdown

Pro plan $21,480

Team plan $15,920

Free → Pro $10,800

Exports

▸demo-en.mp4

▸demo-ja.mp4

▸demo-es.mp4

Filters update the chart live

Rendering exports

✓ demo-en.mp4

✓ demo-ja.mp4

Capabilities

A full studio, exposed as tools.

Everything a human editor would do, callable by your agent or by you.

All features

Agent at the controls

Navigate, click, type, scroll. Your agent operates the browser and your app while the recorder rolls.

Learn more

Cinematic camera

Smooth zooms, pans, and spotlight moves, fired as tool calls or applied by hand in the editor.

Learn more

Self-writing scripts

The agent drafts tight narration from what is actually on screen. Edit a line, regenerate a scene, done.

Learn more

Any voice, your key

Plug in ElevenLabs, OpenAI, or another TTS provider. Pick the voice; pay provider rates directly.

Learn more

One take, every language

Captions and voiceover re-render per language from the same recording. No re-shoots.

Learn more

Local-first recording

Capture and render on your machine. Your footage never touches our servers.

Learn more

Localization

Record once. Ship in every language.

The agent translates the script; your TTS provider voices it. The same demo lands for marketing in Madrid, support in São Paulo, and your team in Tokyo.

demo.mp4

Connect your repo and deploy in one click.

Voiceover: English 1 recording · 8+ exports

Marketing launchesSupport repliesHelp centers & KBsCross-border teams

Bring your own key

Your keys. Your voices. Your costs.

Dope MCP doesn't resell AI minutes. Connect your own ElevenLabs, OpenAI, or other TTS account: use every voice you already have, pay your provider's rates directly, and swap providers anytime. A minute of narration typically costs a few cents, billed by your provider with no markup from us.

Your provider's full voice library
No per-minute markup, ever
Keys stay on your machine

Voice settings

Voice provider

ElevenLabs

API key

sk-••••••••••••••••

Project voices

Amelia

en-GB · warm, clear

Kenji

ja-JP · calm, precise

Usage billed directly by your provider.

Compare

Great tools. Different jobs.

Loom and Cap are excellent at recording yourself. Dope MCP starts where manual recording ends.

Feature	Loom	Cap	Dope MCP
Built for	Quick async video messages	Beautiful manual screen recordings	Agent-produced product demos
Manual recording & editing	Excellent	Excellent	Solid basics
AI agent directs the recording (MCP)
Narration script written by your agent
AI voiceover from a script			BYOK
Multi-language captions + voiceover from one take
Bring-your-own-key AI pricing
Instant share links & viewer analytics	Their strength	Cloud sharing	Exports files instead

Comparison reflects each product's publicly listed features as of August 2026. Loom and Cap are great at what they do; corrections are welcome at hello@dopemcp.com.

Deep dive: Dope MCP vs Loom Deep dive: Dope MCP vs Cap

Pricing

Start free. Upgrade when you ship.

Early-access pricing. Final numbers may change.

Free

For trying agent-directed recording.

Unlimited manual recordings
Full MCP tool set
1080p export
2 languages per project
BYOK text-to-speech

Download free

Pro

Team

$25 /user/month

For teams localizing at scale.

Everything in Pro
Shared brand kits & prompt templates
Shared glossaries for translations
Centralized licenses & key management

Full plan details

FAQ

Questions, answered.

What is MCP?

MCP (Model Context Protocol) is an open standard that lets AI agents call tools in other apps. Dope MCP ships with an MCP server built in, so your agent can operate the recorder the same way it edits your code: navigation, camera, captions, narration.

Which AI agents work with Dope MCP?

Any MCP client: Claude Code, Cursor, Windsurf, Cline, Zed, Continue, GitHub Copilot, Gemini CLI, and more. If your tool can connect to an MCP server, it can direct your videos.

Is my screen recorded locally?

Yes. Capture, editing, and rendering all happen on your machine. The only thing that leaves it is the narration text sent to the TTS provider you configured, with your key, under your account. Nothing is uploaded to us.

What does BYOK text-to-speech cost?

Whatever your provider charges. Dope MCP adds no markup. A one-minute narration is typically a few cents on mainstream TTS APIs, billed directly by ElevenLabs, OpenAI, or whoever you plug in.

Which languages are supported?

Captions and voiceover can be generated in any language your agent can translate to and your TTS provider can speak. For major providers that means dozens of languages from a single recording.

What platforms does it run on?

macOS and Windows during early access.

Do I need an AI agent to use it?

No. Dope MCP works as a normal screen recorder and editor on its own. The agent is the superpower, not a requirement.

Ready to make something dope?

Download the recorder, point your agent at it, and ship your first localized demo today.

Download for macOS Download for Windows Read the docs

Free during early access · No credit card

All platforms & requirements →