Claude + HeyGen: What the Integration Can (and Can't) Do in 2026
HeyGen took MCP seriously, so this is one of the better-supported integrations in this space: HeyGen ships its own official remote MCP server, and Claude can use it to generate real avatar videos. You have three honest routes — HeyGen’s hosted remote MCP server (OAuth, works with Claude web and Claude Code), a local/community MCP server that runs on your own API key, and the plain HeyGen REST API if you’d rather Claude call the endpoints directly.
Here’s each route, when to pick it, and what to use when you want HeyGen videos that get made without you at the keyboard.
Route 1: HeyGen’s official remote MCP server
This is the headline, and it’s the one most people should use. HeyGen hosts an MCP server at https://mcp.heygen.com/mcp/v1/ that supports Claude web, Claude Code, Cursor, and Gemini CLI. Auth is OAuth tied to your HeyGen account — no API key to paste or rotate — and generation spends the premium credits already in your HeyGen plan, with no separate API billing.
In Claude web, you add it under Connectors / MCP servers → Add custom connector, paste the URL, and approve access in the HeyGen sign-in that pops up. In Claude Code, it’s one terminal command (run it in your shell, not inside the Claude Code CLI):
claude mcp add --transport http -s user heygen https://mcp.heygen.com/mcp/v1/Once connected, Claude gets a broad tool set — create_video, get_video, list_videos, list_voices, list_avatar_looks, create_avatar, create_speech, plus the Video Agent tools (create_video_agent, send_video_agent_message) and get_current_user for your remaining credit balance. So you can say “list my avatar looks, then make a 30-second product update in my brand voice” and Claude runs it end to end, polling the status until the render’s done.
Route 2: A local MCP server on your own API key
Before the hosted server, HeyGen’s MCP story started as an open-source local server — the heygen-mcp Python package you run with uvx, wired into Claude Desktop’s config with a HEYGEN_API_KEY from Settings → API in your dashboard. That original project is now effectively superseded by the remote server and is community-maintained through forks, so treat any given fork as third-party code and check what it exposes before trusting it. Functionally the local route covers the same core moves — list voices, generate an avatar video, check render status, read remaining credits.
The config looks like this:
{
"mcpServers": {
"heygen": {
"command": "uvx",
"args": ["heygen-mcp"],
"env": { "HEYGEN_API_KEY": "your-key-here" }
}
}
}Pick this only if you specifically want a local, API-key setup (or you’re already scripting against your key). For everyone else, the OAuth remote server is less to maintain.
Route 3: Claude calling the HeyGen API directly
You don’t strictly need an MCP server. HeyGen’s REST API authenticates with an X-Api-Key header (key from Settings → API), and the useful endpoints are all there: POST https://api.heygen.com/v2/video/generate to create an avatar video from an avatar_id, voice_id, and script, GET /v2/avatars and GET /v2/voices to list what’s available, and a status endpoint to poll the render. (HeyGen’s v1/v2 endpoints are supported through Oct 31, 2026, with v3 as the current generation.) In Claude Code, Claude can read the docs and hit these endpoints for you; in a plain chat, it can write the exact curl or Python. This is the route for building your own automation rather than chatting your way to a clip.
The limits that actually matter
The MCP routes are genuinely good, but they share the ceiling every Claude connector has:
- No triggers. MCP tools only run inside a session you start. A new subscriber, a published blog post, or a Monday morning can’t make Claude generate a video — nothing fires an MCP tool on an event.
- Laptop-bound for anything scheduled. A weekly explainer-video habit means you open Claude, type the prompt, and shepherd the output every single week. Close the laptop and nothing renders.
- Delivery is still on you. Claude hands you a finished video (or a URL). Getting it to your editor, your YouTube channel, or your email list is a separate manual step.
- Local setups don’t reach claude.ai. The Route 2 server runs on your machine over stdio, so the web app can’t see it — that route belongs to Claude Desktop.
Net: this is excellent for making videos on demand in a conversation, and not built for videos that need to happen on a schedule or in response to something.
If you want HeyGen work that runs on its own: Carly
For video that gets produced without you prompting it, that’s the shape Carly is built for — an AI executive assistant that acts on triggers and runs the whole flow in the cloud, laptop closed.
- When a blog post publishes, Carly drafts a script from it, renders a HeyGen avatar video, and emails you the link to approve.
- When a new customer signs up, Carly generates a personalized welcome video and sends it.
- Every Monday at 8am, Carly turns your weekly update into a HeyGen video and posts it where your team looks.
No-code setup. There’s no config file. You say “when I publish a post, make a HeyGen video and email it to me,” Carly interviews you about avatar, voice, and length, and builds the workflow with you.
Actually sends. Carly delivers the result as a real email through Gmail or Outlook — attachment or link — so the job finishes, not just the render.
Connects to anything. Carly reaches HeyGen through your own HeyGen API key (paste it on the integrations page) — one of 200+ native integrations plus anything with an API.
AI agents start at $35/month, and steps in a workflow that don’t use AI run free and unlimited.
Claude vs Carly
| Claude (HeyGen MCP) | Carly | |
|---|---|---|
| List avatars & voices, read your account | Yes | Yes |
| Generate a video when you ask in chat | Yes | Yes |
| Generate a video on a trigger or schedule | No | Yes |
| Works with the laptop closed | No | Yes (cloud) |
| Emails the finished video to someone | No | Yes (Gmail + Outlook) |
| Chains video into a larger workflow | Limited to the session | Yes, as steps |
| Setup | OAuth remote server or local MCP | No-code, Carly interviews you |
| Pricing | Pro $20 / Max $100–$200 | AI agents from $35/mo |
Claude is the best way to make a HeyGen video by talking to it; Carly is the way to make HeyGen videos happen on their own.
Frequently Asked Questions
Does Claude work with HeyGen?
Yes — better than most. HeyGen publishes its own official remote MCP server at mcp.heygen.com/mcp/v1/, so Claude web and Claude Code can generate avatar videos, list voices and avatars, and check render status once you connect it via OAuth.
Is there an official HeyGen MCP server?
There is. HeyGen hosts an OAuth-based remote MCP server that bills against the credits in your existing HeyGen plan. An earlier open-source local heygen-mcp server also exists (now community-maintained), but the hosted one is the supported default.
Can claude.ai on the web access HeyGen?
Yes, using the remote MCP server — add it as a custom connector and approve the HeyGen sign-in. The local, API-key server from Route 2 does not work on the web; it only runs in Claude Desktop.
Can Claude generate a HeyGen video automatically?
Not on its own. MCP tools only run in a session you start, so a publish, signup, or schedule can’t trigger a render. Carly runs the make-and-send flow on triggers 24/7 in the cloud. AI agents start at $35/month.
More: Claude connectors · Claude vs Carly · Can Claude send emails · Claude + ElevenLabs · Claude + Canva · Claude models explained
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
