[Cotlet](https://cotlet.dev/index.md) · [Integrations](https://cotlet.dev/integrations.md) · [How it works](https://cotlet.dev/how-it-works.md) · [Security](https://cotlet.dev/security.md) · [For developers](https://cotlet.dev/for-developers.md) · [For automation agencies](https://cotlet.dev/for-agencies.md) · [Blog](https://cotlet.dev/blog.md) · [API docs](https://cotlet.dev/docs.md) · [Help center](https://cotlet.dev/help.md) · [Changelog](https://cotlet.dev/changelog.md) · [Contact](https://cotlet.dev/contact.md) · [Pricing](https://cotlet.dev/pricing.md) · [Sign up](https://cotlet.dev/signup)

# How to connect CapCut with an AI agent: automate your work

> CapCut publishes no MCP address for Claude or Cursor (checked 6 Nov 2026). Here is how to connect your AI agent to CapCut anyway.

Source: https://cotlet.dev/blog/how-to-connect-capcut-to-chatgpt-claude

CapCut publishes no MCP address for Claude or Cursor (checked 6 Nov 2026). Here is how to connect your AI agent to CapCut anyway.

Cotlet team · Nov 17, 2026 · 15 min read

On this page

1. [01 What your agent does in CapCut: exports, captions, review links and team spaces](#what-your-agent-does-in-capcut-exports-captions-review-links-and-team-spaces)
2. [02 CapCut’s public API, MCP address and ChatGPT plugin as of 6 Nov 2026](#capcuts-public-api-mcp-address-and-chatgpt-plugin-as-of-6-nov-2026)
3. [03 What to check in your CapCut account first: role, plan and export limits](#what-to-check-in-your-capcut-account-first-role-plan-and-export-limits)
4. [04 Step 1: Log in to CapCut in the Cloud Browser and press Enable for agents](#step-1-log-in-to-capcut-in-the-cloud-browser-and-press-enable-for-agents)
5. [05 Step 2: Add the Cotlet.dev address to ChatGPT, Claude or another agent](#step-2-add-the-cotletdev-address-to-chatgpt-claude-or-another-agent)
6. [06 Three CapCut jobs to hand over first: a space audit, a captioned export and a review link](#three-capcut-jobs-to-hand-over-first-a-space-audit-a-captioned-export-and-a-review-link)  
  1. [Find which space has room before you invite another editor](#find-which-space-has-room-before-you-invite-another-editor)
  2. [Caption a draft and export it at 2K](#caption-a-draft-and-export-it-at-2k)
  3. [Send a draft for review and see which links are out](#send-a-draft-for-review-and-see-which-links-are-out)
7. [07 Four other routes into CapCut and what each one covers](#four-other-routes-into-capcut-and-what-each-one-covers)
8. [08 FAQ](#faq)  
  1. [Is there an official CapCut connector for Claude, or only the ChatGPT plugin?](#is-there-an-official-capcut-connector-for-claude-or-only-the-chatgpt-plugin)
  2. [Has CapCut published a public API or an MCP address?](#has-capcut-published-a-public-api-or-an-mcp-address)
  3. [Will Zapier, Make or n8n let an agent run my CapCut drafts?](#will-zapier-make-or-n8n-let-an-agent-run-my-capcut-drafts)
  4. [Can my agent cut clips or add effects to a timeline?](#can-my-agent-cut-clips-or-add-effects-to-a-timeline)
  5. [Do I need CapCut Pro to export through my agent?](#do-i-need-capcut-pro-to-export-through-my-agent)
  6. [Can CapCut tell my agent when a draft changes?](#can-capcut-tell-my-agent-when-a-draft-changes)
  7. [Can I undo what my agent does in CapCut?](#can-i-undo-what-my-agent-does-in-capcut)
  8. [Who can see my CapCut login, and how do I cut access?](#who-can-see-my-capcut-login-and-how-do-i-cut-access)
  9. [What happens when CapCut moves a button or renames a menu?](#what-happens-when-capcut-moves-a-button-or-renames-a-menu)
  10. [Is Cotlet.dev affiliated with CapCut?](#is-cotletdev-affiliated-with-capcut)
9. [09 Run the first space audit on your own CapCut account](#run-the-first-space-audit-on-your-own-capcut-account)

CapCut publishes no MCP address for Claude or Cursor, and its only official AI route is the CapCut plugin for the ChatGPT desktop app on PC (checked 6 Nov 2026). CapCut also has no public API for drafts, exports or sharing (checked 6 Nov 2026), so nothing official lets Claude or another agent work in your account. This guide shows how to connect CapCut to ChatGPT, Claude and any AI agent through Cotlet.dev, which runs 13 CapCut tools in your own account.

CapCut is where a small video team’s weekly work piles up: an owner or manager exports client drafts, adds captions, sends review links and decides who may edit in each team space. Each of those chores repeats for every draft and every client, and the owner usually does them. Cotlet.dev is built for people who run a business in CapCut and want an AI agent to manage their account, while a client who only comments on a cut keeps using CapCut’s own review link.

## What your agent does in CapCut: exports, captions, review links and team spaces

| Area of CapCut | What your agent does there |
| ------------------ | ------------------------------------------------------------------------------------------------------------------------- |
| Team spaces | Lists your spaces with your role and member count, and creates a new space after you approve it. |
| Members and roles | Lists members with their role, reads the invite link, invites people by email and sets a member to admin or collaborator. |
| Drafts | Lists the drafts in a space with who holds editing permission and the last edit. |
| Auto captions | Runs auto captions on a draft and adds the lines to its timeline. |
| Export | Exports a draft with the file name, resolution, quality, frame rate and format you set. |
| Template gallery | Searches the CapCut template gallery for templates that match your search. |
| Review links | Shares a draft through Export > Share for review and lists the drafts shared with you or by you. |
| Editing permission | Hands editing permission on a draft to another team member. |

Your AI agent asks first before anything reaches other people: Share for review, Invite members, Set member role, Transfer editing permission and Create space wait for your OK in the Cotlet.dev console, which covers 5 of the 13 tools. Exports and auto captions reach no one, so they run at once. Cotlet.dev lists the inputs and an example answer of every tool on its [CapCut integration page](https://cotlet.dev/integrations/capcut-api-mcp-server.md).

## CapCut’s public API, MCP address and ChatGPT plugin as of 6 Nov 2026

CapCut has no public API for drafts, exports or sharing and publishes no MCP address for Claude or Cursor (checked 6 Nov 2026). CapCut links no developer portal, no way to get an API key and no webhooks documentation, and Zapier and n8n have no CapCut app page (checked 6 Nov 2026).

CapCut’s official AI route is the [CapCut plugin for ChatGPT](https://www.capcut.com/tools/capcut-x-codex), made for the ChatGPT desktop app on PC and available outside the United States (checked 6 Nov 2026). It generates media, builds an editable timeline and adds captions, effects and templates inside a ChatGPT task. Export, sharing and space management stay in the CapCut editor.

Cotlet.dev fills the gap with one address, <https://cotlet.dev/mcp/v1>, that gives ChatGPT, Claude and other agents 13 CapCut tools. Your AI agent works in your own CapCut account, hosted by Cotlet.dev, so nothing runs on your computer.

## What to check in your CapCut account first: role, plan and export limits

- CapCut login: use your own CapCut account, the one that owns or manages your drafts and spaces, because you sign in to it once in the Cotlet.dev Cloud Browser.
- CapCut role: managing members and roles needs owner or admin rights in the space, so check your role before you ask your agent to invite an editor or set a role.
- CapCut plan: CapCut lists several plans, including Pro, Ultra and Teams, and bills them as before. Cotlet.dev works on the plan you already have.
- CapCut export limits: 2K and 4K depend on your device, the app version and the platform, and the project resolution must be set before editing.
- CapCut space limits: CapCut allows 5 spaces per user, Default space included, and 30 members per space (CapCut help, Aug 2026). Creating a space is permanent, because CapCut does not support deleting one.
- Your AI agent: ChatGPT, Claude or any other agent that accepts an MCP server or connector, with a Cotlet.dev sign-in ready for the approval step.

## Step 1: Log in to CapCut in the Cloud Browser and press Enable for agents

Cotlet.dev gives you a Cloud Browser, a browser that runs on Cotlet.dev’s side and holds your CapCut login. Open the CapCut integration in the Cotlet.dev console, log in to CapCut inside the Cloud Browser with your own account, then press Enable for agents.

CapCut asks for your password on CapCut’s own page. You type it there, and Cotlet.dev never sees or stores it. Cotlet.dev keeps only your logged-in CapCut profile, encrypted with AES-256-GCM in Finland, and your agent gets the 13 CapCut tools, never the login itself.

Cotlet.dev sends your agent a link to sign in again if CapCut signs you out, and you do that sign-in yourself. To cut access, press Disable for agents or reset the browser in the Cotlet.dev console. No tool touches billing, plan, password, two-factor or deleting drafts.

## Step 2: Add the Cotlet.dev address to ChatGPT, Claude or another agent

Cotlet.dev connects CapCut to ChatGPT, Claude, Claude Code, Codex, Cursor, Grok, Hermes Agent, n8n and other MCP clients through one address.

**ChatGPT**chatgpt.com and its apps **Claude**claude.ai and desktop app **Claude Code**one terminal command **Codex**one entry in config.toml **Cursor**one entry in mcp.json **Grok**grok.com connectors **Hermes Agent**one block in config.yaml **n8n**MCP Client node **Any MCP client**OpenClaw and others

Pick your AI above to see its steps.

### How to connect CapCut to ChatGPT

Do this on chatgpt.com in your browser. Cotlet then works in the ChatGPT desktop and phone apps too.

1. Open [Plugins in ChatGPT](https://chatgpt.com/plugins) and click Add.
2. Click Add custom MCP server. Not in the menu? Turn on Developer mode in Settings, Security and login, then come back.
3. Paste the name `Cotlet` and the server URL `https://cotlet.dev/mcp/v1/connect`. Keep Authentication on OAuth.
4. Tick “I understand and want to continue” and click Create as a plugin.
5. Click Connect Cotlet. A Cotlet page opens, click Allow access.
6. In a new chat, type `@Cotlet` and pick it from the list.
7. Ask ChatGPT: “Are you signed in to Cotlet through MCP?” Then try a first CapCut task: “List my CapCut spaces with my role and member count, then show me who can edit in each one and which space has room for another editor.”

_ChatGPT, bottom of the custom MCP server form, November 2026._

_ChatGPT, picking Cotlet in a new chat with @, November 2026._

ChatGPT asks “Allow ChatGPT to use Cotlet?” before some steps. Pick Always allow, or the chat waits for your click. On a Business, Enterprise or Edu workspace, an admin has to allow custom MCP servers first.

### How to connect CapCut to Claude

1. Open [Connectors in Claude](https://claude.ai/customize/connectors) and click Add. In the desktop app it is under Customize, Connectors.
2. Click Add custom connector.
3. Paste the name `Cotlet` and the URL `https://cotlet.dev/mcp/v1/connect`, then click Continue.
4. Keep the settings and click Add.
5. A Cotlet page opens. Click Allow access.
6. In a chat, click the plus button, then Connectors, and turn Cotlet on.
7. Ask Claude: “Are you signed in to Cotlet through MCP?”, then “Add auto captions to the draft Spring Promo in my Client Review space, then export it at 2K and tell me if CapCut refuses the setting.”. Claude asks before it uses a Cotlet tool the first time. Click Allow.

_Claude, Customize, Connectors, Add menu, November 2026._

_Claude, Add custom connector, last screen, November 2026._

Claude’s free plan allows one custom connector. On a Team or Enterprise plan, an owner adds Cotlet in Organization settings, Connectors, and you click Connect next to it.

### How to connect CapCut to Claude Code

Add Cotlet without a key, then sign in once in the browser:

```
claude mcp add --transport http cotlet https://cotlet.dev/mcp/v1/connect
```

Run `/mcp` in Claude Code, pick cotlet and sign in to Cotlet. Prefer an API key, for example on a server? Create one in the Cotlet console and use:

```
claude mcp add --transport http cotlet https://cotlet.dev/mcp/v1 \
  --header "Authorization: Bearer $COTLET_API_KEY"
```

### How to connect CapCut to Codex

Create an API key in the Cotlet console, put it in the `COTLET_API_KEY` environment variable and add this to `~/.codex/config.toml`:

```
[mcp_servers.cotlet]
url = "https://cotlet.dev/mcp/v1"
bearer_token_env_var = "COTLET_API_KEY"
```

### How to connect CapCut to Cursor

Create an API key in the Cotlet console, put it in the `COTLET_API_KEY` environment variable and add this to `~/.cursor/mcp.json`:

```
{
"mcpServers": {
"cotlet": {
"url": "https://cotlet.dev/mcp/v1",
"headers": { "Authorization": "Bearer ${env:COTLET_API_KEY}" }
}
}
}
```

### How to connect CapCut to Grok

Grok takes custom MCP servers on grok.com, per xAI’s [connectors documentation](https://docs.x.ai/grok/connectors) (checked 8 November 2026).

1. Open [grok.com/connectors](https://grok.com/connectors) and click New Connector.
2. Pick Custom and paste the server URL `https://cotlet.dev/mcp/v1/connect`.
3. Sign in to Cotlet when Grok asks, then click Allow access.
4. Ask Grok: “Are you signed in to Cotlet through MCP?”

On Grok Business, an admin adds the connector first in the xAI console, under Grok Business, Connectors. The Grok chat inside X has no custom connector menu.

### How to connect CapCut to Hermes Agent

Add Cotlet to `~/.hermes/config.yaml` in the format of the [Hermes MCP config reference](https://hermes-agent.nousresearch.com/docs/reference/mcp-config-reference):

```
mcp_servers:
cotlet:
url: "https://cotlet.dev/mcp/v1"
headers:
Authorization: "Bearer ${COTLET_API_KEY}"
trust: untrusted
```

Put `COTLET_API_KEY` in `~/.hermes/.env` and check it with `hermes mcp test cotlet`. With `trust: untrusted`, Hermes runs reads at once and asks before every write. More on Hermes in our post on the [Hermes Agent browser](https://cotlet.dev/blog/hermes-agent-browser.md).

### How to connect CapCut to n8n

Use the MCP Client node, or the MCP Client Tool inside an AI Agent node:

- Server Transport: HTTP Streamable
- MCP Endpoint URL: `https://cotlet.dev/mcp/v1`
- Authentication: Bearer Auth, with your Cotlet API key as the token
- Tools: Cotlet.dev’s browser tools; the AI Agent node reads the CapCut guide first

Use it to run this on a schedule: send a draft for review and see which links are out.

### Any other MCP client

Point the client at `https://cotlet.dev/mcp/v1` over Streamable HTTP, with the header `Authorization: Bearer <your Cotlet API key>`. A client that supports OAuth sign-in can use `https://cotlet.dev/mcp/v1/connect` instead and needs no key. The full reference is in the [MCP docs](https://cotlet.dev/docs/mcp.md).

1. Your AI agent receives the address: add <https://cotlet.dev/mcp/v1> as an MCP server or connector in its settings.
2. Your AI agent then asks you to sign in to Cotlet.dev, and you approve the connection.
3. Your AI agent is connected once a small request such as “List my CapCut spaces with my role and member count” returns your spaces.

Cotlet.dev uses that one shared address for CapCut and your other apps. If you enable CapCut after connecting your agent, refresh the connector so the new tools show up. ChatGPT and Claude then work in your CapCut account from their phone apps too.

## Three CapCut jobs to hand over first: a space audit, a captioned export and a review link

### Find which space has room before you invite another editor

Your AI agent takes this request: “List my CapCut spaces with my role and member count, then show me who can edit in each one and which space has room for another editor.” The agent only reads, so nothing waits for approval. CapCut allows 5 spaces per user, Default space included, and 30 members per space ([CapCut help, Aug 2026](https://www.capcut.com/resource/collaborate-on-capcut-online)), and the agent checks every space against those limits. It replaces opening each space and counting members by hand.

### Caption a draft and export it at 2K

Your AI agent takes this request: “Add auto captions to the draft Spring Promo in my Client Review space, then export it at 2K and tell me if CapCut refuses the setting.” The agent finds the draft in the space you name, runs auto captions, adds the lines to the timeline, then exports with the resolution, quality, frame rate and format you set. Neither step reaches anyone, so both run at once and the file lands in the Cloud Browser’s downloads. Because 2K and 4K depend on your device and app version, a setting CapCut refuses fails with CapCut’s own message and changes nothing.

### Send a draft for review and see which links are out

Your AI agent takes this request: “Share the draft Spring Promo for review with everyone who has the link, then list every review link I have shared.” The agent looks the draft up by name, sets the visibility to everyone with the link or to the space only, and waits: the share shows up in your Cotlet.dev console first, and you approve it there. After your OK it reports the review link, the visibility and who got an email. Cotlet.dev has no revoke tool for review links, so the share history is how you see which links are out, and it saves you the trip through Export > Share for review for every old draft.

## Four other routes into CapCut and what each one covers

- CapCut’s plugin for ChatGPT is official, made for the ChatGPT desktop app on PC, and builds and edits drafts inside a ChatGPT task ([CapCut × ChatGPT plugin](https://www.capcut.com/tools/capcut-x-codex), checked 6 Nov 2026).
- CapCut has no public API for drafts, exports or sharing and no MCP address for Claude or Cursor, so no tool can call one (checked 6 Nov 2026).
- CapCut has no app page on Zapier or n8n (checked 6 Nov 2026), and it documents no webhooks for them to listen to.
- CapCut community servers on GitHub, such as [VectCutAPI](https://github.com/sun-guannan/VectCutAPI), write CapCut and Jianying draft files on your own computer with Python and FFmpeg. The project has 2,362 stars and its last commit is 24 Oct 2026, and you open the file in CapCut and run the export there.

Cotlet.dev adds what none of these give: your own signed-in CapCut account, with the drafts, exports, review links and team spaces you manage, read and write, from ChatGPT, Claude or any agent, with no API key, on the CapCut plan you already have.

## FAQ

### Is there an official CapCut connector for Claude, or only the ChatGPT plugin?

CapCut has one official AI route, the CapCut plugin for the ChatGPT desktop app on PC, and no connector for Claude (checked 6 Nov 2026). The plugin builds and edits drafts inside a ChatGPT task and is available outside the United States. For Claude, add <https://cotlet.dev/mcp/v1> as a connector and your agent works in your CapCut account.

### Has CapCut published a public API or an MCP address?

CapCut has no public API for drafts, exports or sharing and publishes no MCP address for Claude or Cursor (checked 6 Nov 2026). It links no developer portal and no webhooks documentation. Cotlet.dev runs 13 CapCut tools in your own account over MCP, so your agent needs neither.

### Will Zapier, Make or n8n let an agent run my CapCut drafts?

CapCut has no app page on Zapier or n8n (checked 6 Nov 2026), and no public API for Make or any other tool to call. Cotlet.dev instead gives n8n and other MCP clients the 13 CapCut tools through <https://cotlet.dev/mcp/v1>.

### Can my agent cut clips or add effects to a timeline?

Cotlet.dev’s CapCut tools cover drafts, auto captions, export, review links, team spaces and template search, not cutting, effects or transitions. Cutting stays in the CapCut editor, where you or your team work by hand. For building a timeline in chat, CapCut’s own plugin for ChatGPT adds captions, effects and templates.

### Do I need CapCut Pro to export through my agent?

CapCut’s help says 2K and 4K depend on your device, the app version and the platform, and the project resolution must be set before editing. When CapCut refuses a setting or asks for a plan, the export tool reports it and changes nothing else. Cotlet.dev does not change your CapCut plan, which CapCut bills as before.

### Can CapCut tell my agent when a draft changes?

CapCut documents no webhooks (checked 6 Nov 2026), so nothing tells an agent that a draft changed. Run your agent on a schedule instead, for example a weekly pass over the draft list and the share history.

### Can I undo what my agent does in CapCut?

Cotlet.dev’s CapCut tools undo differently by action. An export is undone by deleting the file, since the draft is unchanged, and auto captions by deleting the caption lines in the editor. A role change is set back by setting the old role, and a transfer of editing permission is taken back the same way. Invites and review shares cannot be taken back by a tool, and a created space is permanent, which is why those wait for your OK.

### Who can see my CapCut login, and how do I cut access?

Cotlet.dev stores your logged-in CapCut profile, encrypted with AES-256-GCM in Finland, and not your password. Your agent gets the 13 CapCut tools and never the login itself. To cut access, press Disable for agents or reset the browser in the Cotlet.dev console.

### What happens when CapCut moves a button or renames a menu?

Cotlet.dev tests the CapCut tools every 12 h, and the fix happens on its side when CapCut changes a menu, so your prompts keep working. If CapCut signs you out, your agent gets a link to sign in again, and you do that step yourself.

### Is Cotlet.dev affiliated with CapCut?

Cotlet.dev is independent and not affiliated with CapCut or ByteDance. CapCut’s own plugin remains the official route inside the ChatGPT desktop app.

## Run the first space audit on your own CapCut account

Your AI agent is ready for work in CapCut as soon as the connector lists the 13 tools. Start with your busiest team space: ask for the audit above, then approve each review share and each invite yourself. Cotlet.dev lists the inputs and an example answer of every tool on its [CapCut integration page](https://cotlet.dev/integrations/capcut-api-mcp-server.md).

## Keep reading

[All posts](https://cotlet.dev/blog.md)

- [Agents Nov 18, 2026 How to connect Petpooja with an AI agent: automate your work Petpooja has no official MCP server (checked 31 Oct 2026). Connect Petpooja to ChatGPT, Claude or any AI agent with Cotlet.dev in ten minutes. Cotlet team · 16 min read](https://cotlet.dev/blog/how-to-connect-petpooja-to-chatgpt-claude.md)
- [Agents Nov 18, 2026 How to connect Systeme.io to AI agents and automate the busywork Systeme.io has an official MCP server but it skips refunds, invoices and blog posts. Connect ChatGPT, Claude or any AI agent with Cotlet.dev. Cotlet team · 15 min read](https://cotlet.dev/blog/how-to-connect-systeme-io-to-chatgpt-claude.md)
- [Agents Nov 18, 2026 How to connect Wanderlog to ChatGPT and Claude, no code needed Wanderlog has no official MCP server (checked 24 Oct 2026). Connect Wanderlog to ChatGPT, Claude or any AI agent with Cotlet.dev in two steps. Cotlet team · 15 min read](https://cotlet.dev/blog/how-to-connect-wanderlog-to-chatgpt-claude.md)

## Have an app in mind?

Sign up and add it in the console. Not in the list? Ask for it there.

[Sign up free](https://cotlet.dev/signup)

---

Give your AI agents your signed-in browser.

**Product:** [Integrations](https://cotlet.dev/integrations.md) · [How it works](https://cotlet.dev/how-it-works.md) · [Pricing](https://cotlet.dev/pricing.md) · [Security](https://cotlet.dev/security.md)

**Resources:** [Blog](https://cotlet.dev/blog.md) · [API docs](https://cotlet.dev/docs.md) · [Help center](https://cotlet.dev/help.md) · [Changelog](https://cotlet.dev/changelog.md) · [GitHub](https://github.com/cotletdev/cotlet)

**Solutions:** [For developers](https://cotlet.dev/for-developers.md) · [For automation agencies](https://cotlet.dev/for-agencies.md)

**Company:** [About](https://cotlet.dev/about.md) · [AI info: company facts](https://cotlet.dev/ai-info.md) · [Contact](https://cotlet.dev/contact.md)

**Legal:** [Terms of service](https://cotlet.dev/terms) · [Privacy policy](https://cotlet.dev/privacy) · [Refund policy](https://cotlet.dev/refunds) · [Acceptable use](https://cotlet.dev/acceptable-use) · [Cookie policy](https://cotlet.dev/cookies) · [DPA](https://cotlet.dev/dpa) · [Subprocessors](https://cotlet.dev/subprocessors)

**Elsewhere:** [Status page](https://status.cotlet.dev) · [Cotlet on GitHub](https://github.com/cotletdev) · [Cotlet on X](https://x.com/cotletdev) · [Cotlet on LinkedIn](https://www.linkedin.com/company/cotletdev) · [Cotlet on Product Hunt](https://www.producthunt.com/products/cotlet) · [Cotlet on Crunchbase](https://www.crunchbase.com/organization/cotlet) · [Cotlet on SaaSHub](https://www.saashub.com/cotlet)

Cotlet.dev is built by Bart Krawczyk and Jake Slugocki, the founders of Take Group sp. z o.o., ul. Wincentego Janasa 4, 40-855 Katowice, Poland. KRS 0000871398, NIP 6342984273, VAT EU: PL6342984273. [About us](https://cotlet.dev/about.md)

Cotlet is an independent service. It is not affiliated with, endorsed by or sponsored by any of the products listed on this site. Product names are trademarks of their respective owners.
