VlogMe MCP server for AI agents

The MCP server at https://vlogme.ai/mcp gives an AI agent the same work as the VlogMe app, on your account, your credit balance and your Lab:

• A finished video from an idea. The Director plans the scenes, writes the lines, prepares pictures, voices and music, and renders the video with subtitles.
• Single clips and pictures. Text-to-video and image-to-video models, talking photos, avatars, lip sync, upscaling and image generation; the catalog lists each model with its limits and price.
• Voices and sound. Voice-overs, voice changes, music, sound effects, transcripts and captions; dubbing into other languages.
• Edits of your own footage with Aisma, and News Videos from articles you choose.
• Your files. Upload a local file, import one from a public link, list and rename Lab files, and get download links for the results.

Every result is saved to your Lab, where you can open it in the app. There are 59 tools in all; the REST API reference lists each one with the key scope it needs.

Only the jobs you start cost credits: renders and other generations.

At no charge: connecting, the tool list, the job catalog, price quotes, reading your files and jobs, and the Director's planning, chat and free preparation. The planning and preparation share the free allowance your account has in the app; when it is used up for now, the agent's answer says limited: true.

Credits: renders, the same as in the app. An agent pays exactly what you would pay on the website, from the same balance:

1. The agent asks for a quote. Nothing is charged.
2. It shows you the credits and waits for your yes.
3. It starts the job with that quote. A job never runs at another price: if the price or the input changed, the start is refused and nothing is charged.

A job that fails is not charged. Credits are bought on the website, in Account; a key cannot buy them. When the balance is short, the agent gets a refusal that says so and links to the page where you add credits. See pricing for packs and plans.

Sign in, no key to copy. Clients that support MCP sign-in (Claude Code, claude.ai and ChatGPT connectors, Cursor, VS Code) need only the URL https://vlogme.ai/mcp. The client opens VlogMe in your browser: you sign in, choose what the agent may do and press Allow. The app then appears in Account → API & MCP → Connected apps, where you can disconnect it at any time.

Or use an API key for scripts, n8n and clients without sign-in:

1. Create a key in Account → API & MCP (the account area is in English). Copy it once: it starts with vma_.
2. Keep it in an environment variable, so it never lands in a file you share:

export VLOGME_API_KEY="vma_…"

3. Add the server to your client with these values:

The sections below show the exact steps for each client. To load all 59 tools as separate entries in your client, add ?tools=all to the URL: https://vlogme.ai/mcp?tools=all.

Anyone can read the job catalog (https://vlogme.ai/api/catalog) and the OpenAPI document (https://vlogme.ai/api/openapi.json) without a key, so an agent can plan before you connect it.

With sign-in: run claude mcp add --transport http vlogme https://vlogme.ai/mcp, then type /mcp in Claude Code, pick vlogme and choose Authenticate. Your browser opens VlogMe; press Allow.

With a key: Run this once in a terminal where VLOGME_API_KEY is set:

claude mcp add --transport http vlogme https://vlogme.ai/mcp --header "Authorization: Bearer $VLOGME_API_KEY"

Check it with claude mcp list, then ask: “Use VlogMe to make a 5-second clip of a paper boat on a pond. Quote it first.” Claude Code asks you before each tool call unless you allowed that tool.

Add the server to ~/.cursor/mcp.json (or .cursor/mcp.json in a project). Cursor reads the key from your environment:

{
  "mcpServers": {
    "vlogme": {
      "url": "https://vlogme.ai/mcp",
      "headers": { "Authorization": "Bearer ${env:VLOGME_API_KEY}" }
    }
  }
}

Open Cursor Settings → MCP to see the server and its tools. Start Cursor from a terminal where the variable is set if the tools do not appear.

Add .vscode/mcp.json to your workspace. VS Code asks for the key the first time and keeps it in its secret storage:

{
  "inputs": [
    { "type": "promptString", "id": "vlogme-key", "description": "VlogMe API key", "password": true }
  ],
  "servers": {
    "vlogme": {
      "type": "http",
      "url": "https://vlogme.ai/mcp",
      "headers": { "Authorization": "Bearer ${input:vlogme-key}" }
    }
  }
}

Start the server from the file or the MCP view, then use the tools in agent mode in Chat.

Run this once. The single quotes store a reference to VLOGME_API_KEY in your user settings, not the key itself; Gemini CLI reads the variable when it starts:

gemini mcp add --scope user --transport http vlogme https://vlogme.ai/mcp --header 'Authorization: Bearer $VLOGME_API_KEY'

Or add the server to ~/.gemini/settings.json yourself:

{
  "mcpServers": {
    "vlogme": {
      "httpUrl": "https://vlogme.ai/mcp",
      "headers": { "Authorization": "Bearer $VLOGME_API_KEY" }
    }
  }
}

In Gemini CLI, /mcp lists the server and its tools.

Add the server to ~/.codex/config.toml. Codex sends the key from the variable you name:

[mcp_servers.vlogme]
url = "https://vlogme.ai/mcp"
bearer_token_env_var = "VLOGME_API_KEY"
tool_timeout_sec = 120

tool_timeout_sec gives the tools that wait for a job (up to 30 seconds a call) room to answer.

In a workflow with an AI Agent node, add an MCP Client Tool and set:

• Endpoint: https://vlogme.ai/mcp
• Server Transport: HTTP Streamable
• Authentication: Bearer Auth, with a credential that holds your key

The agent can then quote jobs, start them and wait for them. Every paid start still needs a fresh quote and its fingerprint, so the workflow always pays the quoted price. For a daily video, let the workflow compare the quoted credits with a limit you set before it starts the job.

A key does what you ticked when you made it, and nothing more:

• read (every key): the catalog, quotes, your files, jobs and balance. A read key is enough to let an agent explore and price work.
• write: upload and import files, plan with the Director and start paid jobs at their quoted price.
• delete, share and publish: deleting files and projects, public share links, and posts on your connected social accounts. Leave them off unless the agent needs them.

A key can never buy credits, change your plan, make other keys or delete your account. It expires after 1 to 365 days, and revoking it in Account → API & MCP stops every client that uses it at once. Keep it secret: it acts for your account (Terms of Service).

Connected apps. An app connected by sign-in gets the abilities you ticked on the Allow page, from the same list. Its access renews itself and stops at once when you disconnect it in Account → API & MCP → Connected apps.

Try one of these prompts after you connect:

• “With VlogMe, make a 30-second explainer about how solar panels work. Show me the plan and the price before you render.”
• “Animate ./photo.jpg into a 5-second clip with a slow zoom. Quote it first.”
• “Make my portrait talk: ./me.png says ‘Welcome to our shop’. Which model fits, and what does it cost?”
• “What can VlogMe make, and what does a 10-second video cost?”

Behind the scenes the agent follows the same steps every time: it reads the catalog, asks for a quote, shows you the credits, starts the job with that quote and a new idempotency key, waits for the job, and gives you a download link that lasts 15 minutes. A Director video adds planning and preparation before its quote. The MCP section of the REST reference describes each step.

Is the MCP server free? Connecting, the catalog, quotes and Director's planning and preparation cost nothing, within the same usage limits as the app. The jobs you start, renders included, cost the same credits as in the app.

Can an agent spend my credits without asking? Every paid tool tells the agent to show you the quoted credits and wait for your yes, and your client asks before each tool call unless you allowed it. A job only starts at the price of its quote. To let an agent explore without any spending, give it a key with only read.

Does an agent pay more than the website? No. The prices, the balance and the free allowance are the same.

Do results carry a watermark? VlogMe adds no visible watermark to what you make.

Which clients work? Any MCP client that can send an Authorization header over Streamable HTTP, including Claude Code, Cursor, VS Code, Gemini CLI, Codex and n8n. Clients that connect through an OAuth sign-in, such as claude.ai and ChatGPT connectors, need no key: they sign you in and you press Allow.

How many tools will my agent see? Up to 59. Add ?tools=all to the URL to load every tool as its own entry. Without it, the server can answer a shorter core list, and vlogme_find_tools and vlogme_call_tool then reach the rest. The two paid tools are always in the short list, so your client confirms each call. Tools that delete, share or publish are never run through vlogme_call_tool; to use them, connect with ?tools=all.

Where are my files? In your Lab, the same as anything you make in the app. Download links in tool answers last 15 minutes; ask again for a new one.

What if a request is refused? The tool answer says why, in plain words, and nothing is charged for a refused call. Too many requests in a short time are refused with rate_limited and the seconds to wait.