Use Vishot from AI agents (MCP)

Claude, ChatGPT, Cursor, GitHub Copilot, Gemini CLI, Codex and other AI agents can work in your Vishot account through MCP (Model Context Protocol): check your plan and balance, create sets and dialogue scenes, start video generation and fetch the result. The agent works through your account and your keys, with the same plan limits as the plugin.

The server address

https://vishot.ai/api/mcp

Add this address in the agent. The first time, the agent opens a Vishot page in your browser: sign in, check the app name and press Allow. You never paste a password or a key into the agent. You can see and disconnect every connected agent in Account › AI agents.

What an agent can do

Tool	What it does	Costs money
`get_account`	plan, credit balance, video seconds left this month	no
`list_video_models`	cloud video models with price per second	no
`estimate_video`	price of one video before it starts	no
`list_layouts`, `get_layout`	your sets, their characters, marks, cameras and scenes	no
`create_layout`	a new set from a template, such as a room, a village street or a forest clearing	no
`get_scene`, `create_scene`, `update_scene`	dialogue scenes and the shot plan made from them	no
`generate_video`	starts one video in the cloud and returns its id	yes
`get_video`, `list_videos`	status, progress and the download link	no
`cancel_video`	cancels a video still in the queue; the credit comes back	no

Creating sets and creating or changing scenes through an agent needs a paid plan, like the full web editor. On the free plan an agent can still read your account, sets, scenes and videos, and Studio in the browser opens with the sample scene.

Paid generation is always capped

generate_video runs on your own provider key if you connected one in Providers and keys, otherwise on platform credit. Every call carries max_cost_usd: if the video would cost more, nothing starts. Agents are told to estimate first and tell you the price. Most agents also ask before each tool call; keep that on for generate_video.

When you connect, you choose what the agent may do: read, scenes, paid generation. An agent without paid generation cannot spend anything.

Paid generation is off until you tick it on the consent page. Each agent with paid generation has its own monthly limit in USD: $5 unless you set another amount there (from $1 to $1000). When this month's videos of the agent plus the next max_cost_usd would go over the limit, the video does not start. You can change the limit in My account › AI agents.

Claude (claude.ai, Desktop, mobile)

Open Settings › Connectors and press Add custom connector.
Name it Vishot, paste https://vishot.ai/api/mcp and press Add.
Press Connect, sign in to Vishot and press Allow.
In a chat, turn Vishot on in the tools menu and ask, for example: "Make a two-person scene in a room set and estimate a 6-second video of the first line."

On Team and Enterprise plans an owner adds the connector for the organization first.

Claude Code

claude mcp add --transport http vishot https://vishot.ai/api/mcp

Then run /mcp, choose vishot and Authenticate.

ChatGPT

Open Settings › Apps (on some plans Connectors), then Advanced settings, and turn on Developer mode.
Press Create app, name it Vishot, paste https://vishot.ai/api/mcp as the server URL and choose OAuth authentication.
Confirm, sign in to Vishot and press Allow.
In a new chat, pick Vishot from the tools menu.

Developer mode is available on paid ChatGPT plans; in Business and Enterprise workspaces an admin may need to allow it.

OpenAI Codex

codex mcp add vishot --url https://vishot.ai/api/mcp
codex mcp login vishot

Cursor

Add to ~/.cursor/mcp.json (all projects) or .cursor/mcp.json (one project):

{
  "mcpServers": {
    "vishot": { "url": "https://vishot.ai/api/mcp" }
  }
}

Open Cursor Settings › MCP, find vishot and press Needs login.

GitHub Copilot in VS Code

Add to .vscode/mcp.json, or run MCP: Add Server from the Command Palette and choose HTTP:

{
  "servers": {
    "vishot": { "type": "http", "url": "https://vishot.ai/api/mcp" }
  }
}

Press Start above the server, then sign in when VS Code asks. Use Copilot Chat in Agent mode.

Gemini CLI

gemini mcp add --transport http vishot https://vishot.ai/api/mcp

Start gemini and run /mcp auth vishot.

Windsurf and other editors

Most editors take the same address. Windsurf: add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "vishot": { "serverUrl": "https://vishot.ai/api/mcp" }
  }
}

If an app cannot sign in with OAuth, use a personal key (below).

Personal keys: scripts, n8n and model APIs

For tools that cannot open a sign-in page, create a key in Account › AI agents › Personal key. It is shown once. A key expires after 30, 90 (the default) or 365 days, or never if you choose so; a key with paid generation gets its own monthly limit, like an agent. A key that has not been used for 180 days is revoked automatically. Send it as a header:

Authorization: Bearer vsk_...

OpenAI Responses API:

"tools": [{
  "type": "mcp",
  "server_label": "vishot",
  "server_url": "https://vishot.ai/api/mcp",
  "headers": { "Authorization": "Bearer vsk_..." },
  "require_approval": "always"
}]

Claude API (MCP connector, beta header anthropic-beta: mcp-client-2025-11-20):

"mcp_servers": [{ "type": "url", "name": "vishot", "url": "https://vishot.ai/api/mcp", "authorization_token": "vsk_..." }],
"tools": [{ "type": "mcp_toolset", "mcp_server_name": "vishot" }]

Give a key only the permissions it needs, and revoke it in Account › AI agents if it leaks.

Privacy

The agent sees what the tools return: your plan, balance, sets, scenes and video links. Vishot does not see your chat with the agent. Prompts and images you pass for a video go to the video provider the same way as from the plugin and are not stored on the portal; reference images from your Studio library stay in your account. More in Privacy and data.