PopNow MCP Server

The PopNow MCP Server is a hosted Model Context Protocol server at https://api.popnow.ai/mcp that lets ChatGPT, Claude, Cursor, VS Code and any other MCP client create AI images, videos, songs and voiceovers on PopNow, using your PopNow credits.

Ask in plain words ("turn this photo into a 5-second video", "make an upbeat synth-pop track") and the assistant picks the right PopNow tool, tells you the price, and returns the result in the chat. There is nothing to host: the server is remote, signs you in with OAuth 2.1, and uses the same models, prices and library as popnow.ai. It is also the home of PopNow's free Agent Skills: ask for a style for your launch video, deck or landing page and the assistant shows you presets to pick from. Make it pop.

This repository contains the public setup guide, copy-paste client configs in examples/ and an MCP registry manifest.

Building with an API key (Claude Code, Cursor, Codex or your own code, billed in USD from an API balance)? Use the developer server in popnow-api-mcp. The @popnow/mcp npm bridge moved there too.

Contents

• Quickstart

• Tools

• Style presets: find_skills and use_skill

• Connect your client: Claude · Claude Desktop (local) · Claude Code · ChatGPT · Cursor · VS Code · Windsurf · Cline · Codex CLI · Other clients

• Hermes Agent and Gemini CLI extensions

• Authentication

• Credits, prices and confirmation

• Example prompts

• Troubleshooting

• FAQ

Related MCP server: media-gen-mcp

Quickstart

1. Copy the server URL: https://api.popnow.ai/mcp

2. Add it to your client as a remote (Streamable HTTP) MCP server. For example, in Claude Code:

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

3. Sign in and ask. Your client opens a PopNow sign-in page the first time; press Allow, then ask for a picture, a video or a song.

No PopNow account yet? Sign up at popnow.ai and pick a credit pack when you need more credits.

Tools

• Models. Every generation tool that takes model accepts a PopNow model id from list_models, so "use Seedance" or "make it with Wan" just works. Models include Kling, Seedance, Wan, Vidu, MiniMax, Nano Banana, GPT Image, Flux, Suno and more; the live list is always list_models.

• Photos in. image_url takes a public https link or a popnow:// link from an earlier result. In ChatGPT, an attached photo arrives as the image file object.

• Results out. Images come back inline as a preview plus a download link; videos and audio come back as links. Each result also has a permanent popnow:// link you can pass to another tool ("now animate that picture"). Everything is saved in your PopNow library.

• Annotations. All tools carry the standard MCP hints. get_job, list_my_creations, check_balance, list_models, find_skills and use_skill are read-only, and no tool deletes anything.

• Effects cover the same set as the Effects page. Two-person effects need two placed photos and are website-only.

Style presets: find_skills and use_skill

PopNow keeps a catalog of free Agent Skills: named styles for product launch videos, slide decks, landing pages and websites, editorial images and more, each with a preview and a full playbook that an AI agent follows to build the result. Two free tools bring the catalog into any MCP client:

1. Find. The assistant calls find_skills with what you want to make, in your words (optionally type: video, image, slides or web). It gets up to 6 options (4 by default), best first, each with a name, a one-line description of the look, a preview link, the PopNow generation it uses and its local requirements.

2. Pick. The assistant shows you the options with their previews and asks which one you want.

3. Use. The assistant calls use_skill with the chosen slug and agent (claude-code, codex, cursor, gemini-cli, github-copilot, opencode, or chat in apps without a file system). The reply holds the full SKILL.md playbook, a short-lived signed download link to the complete package with the install commands for that agent, and which PopNow tools (create_image, create_video, create_music, create_speech ...) replace the playbook's own generation scripts.

Finding, downloading and using a preset is free. Only the images, video, music or voice it generates cost credits, quoted and confirmed like every other PopNow tool. Presets are licensed for your own and client work; please don't republish the packages. In a chat app without a file system the assistant can still follow the playbook's look and make the pictures with create_image; steps that run local scripts need a coding agent.

You:        I need a launch video for our note-taking app.
Assistant:  (calls find_skills with intent "launch video for a note-taking app", type "video")
            Here are three styles:
            1. <Preset A> - <its one-line look>. Preview: <link>. Uses PopNow video; needs Node.js.
            2. <Preset B> - ...
            3. <Preset C> - ...
            Which one do you like? The preset is free; only the clips it makes cost credits.
You:        The first one.
Assistant:  (calls use_skill with that slug and agent "claude-code", installs the package from the
            signed link, then follows the playbook)
            Installed. The first clip will cost 185 credits. Go ahead?
You:        Yes.
Assistant:  (calls create_video with max_credits: 185, then get_job, and keeps building the video)

You can also browse the catalog and its previews on the website; Use a skill in ChatGPT or Claude walks through it.

Connect your client

Every client below talks to the same remote server and signs in with your PopNow account (OAuth). Ready-made files are in examples/.

Claude (web, desktop and mobile)

Custom connectors are available on Claude plans and workspaces that allow them.

1. Open Settings → Connectors (claude.ai/settings/connectors) in Claude on the web or in the desktop app.

2. Click Add custom connector, name it PopNow and paste https://api.popnow.ai/mcp. Leave the advanced OAuth fields empty: Claude registers itself automatically.

3. Click Connect, sign in to PopNow (Google, GitHub or an email code) and press Allow.

A connector added on the web also works in the Claude mobile apps. If you do not see Connectors, ask your workspace admin.

Claude Desktop (local config)

Prefer the connector above. If you want PopNow in claude_desktop_config.json instead (macOS: ~/Library/Application Support/Claude/claude_desktop_config.json, Windows: %APPDATA%\Claude\claude_desktop_config.json), use the generic mcp-remote stdio bridge, which opens the PopNow sign-in page the first time (examples/claude-desktop-mcp-remote.json):

{
  "mcpServers": {
    "popnow": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://api.popnow.ai/mcp"]
    }
  }
}

Restart Claude Desktop after editing the file. mcp-remote needs Node.js 18 or newer.

Claude Code

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

Then run /mcp inside Claude Code and choose PopNow to sign in.

Add --scope user to make PopNow available in every project, or --scope project to share it through .mcp.json.

ChatGPT

Custom MCP servers depend on your ChatGPT plan and workspace settings, and menu names change as ChatGPT updates. The current path:

1. Go to chatgpt.com/plugins, press + and choose Add custom MCP server.

2. Name it PopNow, paste https://api.popnow.ai/mcp under Connection and pick OAuth as the sign-in method. Accept the warning, then choose Create as a plugin.

3. Sign in to PopNow and press Allow.

4. In a new chat, type @, pick PopNow and ask. Photos you attach are passed to PopNow directly.

Cursor

Add to .cursor/mcp.json in your project, or ~/.cursor/mcp.json for all projects (examples/cursor.mcp.json):

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

Cursor starts the OAuth sign-in when it connects.

VS Code

Add to .vscode/mcp.json (note the top-level key is servers) (examples/vscode.mcp.json):

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

VS Code opens the sign-in page the first time the server starts.

Windsurf

Open the MCP settings in Windsurf and edit the raw mcp_config.json (its location is in the official Windsurf MCP docs). Remote servers use serverUrl (examples/windsurf.mcp_config.json):

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

Press Refresh in the MCP panel after saving, then sign in when Windsurf asks.

Cline

Open MCP Servers → Configure and add PopNow to cline_mcp_settings.json. Set type to streamableHttp explicitly; without it Cline falls back to the legacy SSE transport (examples/cline_mcp_settings.json):

{
  "mcpServers": {
    "popnow": {
      "type": "streamableHttp",
      "url": "https://api.popnow.ai/mcp",
      "disabled": false,
      "autoApprove": []
    }
  }
}

Use Authorize OAuth in Cline (or in the cline mcp wizard of the Cline CLI) to sign in.

To add the hosted server directly from Cline CLI:

cline mcp add popnow https://api.popnow.ai/mcp --transport http --yes

The same steps are summarized in llms-install.md. Once signed in, call check_balance with {} to check the connection without spending credits.

Hermes Agent and Gemini CLI extensions

This repository is also a portable Agent Plugins v1 package. It declares the hosted server in mcp.json; it does not start a local executable or require the optional npm bridge.

For Hermes Agent, use its native MCP configuration so OAuth is explicit. Add to the active profile's config.yaml:

mcp_servers:
  popnow:
    url: https://api.popnow.ai/mcp
    auth: oauth
    trust: untrusted

Run hermes mcp login popnow, then start a new session or use /reload-mcp. The native configuration asks Hermes to require approval before write-capable calls, including generation. PopNow also enforces its returned price and confirmation requirements on the server. Reading a balance or listing models does not spend credits.

For Gemini CLI:

gemini extensions install https://github.com/popnow-ai/popnow-mcp

The root gemini-extension.json uses Streamable HTTP and OAuth. Run /mcp auth popnow in Gemini CLI to sign in if needed. Tool approval follows the client's normal policy.

Data and execution disclosure. These manifests connect to https://api.popnow.ai/mcp. Only the tool arguments and media references sent by your client are passed to PopNow. Generation creates media in the connected account and spends its prepaid balance. The portable package has no hooks, local commands, background processes, or custom credential storage. Review the Privacy Policy and Terms, or contact support@popnow.ai.

Marketplace review is separate from direct installation. These installation commands do not imply approval or endorsement by a client vendor.

Codex CLI

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

Or add it to ~/.codex/config.toml (examples/codex.config.toml) and run codex mcp login popnow:

[mcp_servers.popnow]
url = "https://api.popnow.ai/mcp"

Other clients

Any client that speaks Streamable HTTP and supports OAuth sign-in can use the same URL (examples/generic.json). Clients that only run local stdio servers can use mcp-remote, as in the Claude Desktop config.

To call PopNow from your own code, scripts or CI with an API key, use the developer server in popnow-api-mcp or the PopNow REST API.

Authentication

OAuth 2.1 (default). PopNow is a standard OAuth 2.1 authorization server: authorization code with PKCE (S256), refresh tokens, dynamic client registration (RFC 7591) and client ID metadata documents. Discovery starts from the 401 on /mcp, whose WWW-Authenticate header points to /.well-known/oauth-protected-resource/mcp; server metadata is at /.well-known/oauth-authorization-server. The consent page shows the app's name and the two scopes, generate and read. Access tokens last one hour and refresh automatically; a connection stays signed in until it goes unused for 90 days. The AI app never sees your password or payment details.

API keys are for developers. They are billed in USD from a separate API balance and connect to the developer server at https://api.popnow.ai/api/v1/mcp; see popnow-api-mcp. This address still accepts API keys until 31 January 2027 so existing setups keep working while they move.

Disconnect at any time in Account settings → Devices → Connected AI apps on popnow.ai.

Credits, prices and confirmation

Connecting is free, and so are find_skills and use_skill. Every creation spends the same credits as on popnow.ai, at the same prices, from the same balance: 1 credit = $0.001. PopNow sells credit packs with no subscription, from Starter ($9, 9,000 credits) to Pro ($99, 120,000 credits including bonus), or any amount from $10 to $2,000 at 1,000 credits per $1; see pricing.

Every generation tool quotes first:

• Up to 300 credits (an image, a song, a 5-second video on the default model): the job starts right away and the reply states what it cost.

• Over 300 credits (longer or higher-resolution video, some effects): nothing starts. The tool answers Needs confirmation with the price and your balance, the assistant asks you, and when you agree it calls the tool again with max_credits set to the price you accepted.

• Price moved between quote and start: nothing is charged and the reply gives the new price to confirm.

• A creation fails: its credits are refunded automatically.

Identical calls from the same account within ten seconds create one job, so an assistant that retries does not pay twice. A tool waits up to about 40 seconds for the result; longer jobs (most videos) return a job_id and the assistant collects the result with get_job.

A creation over the limit is not charged; the tool says how many seconds to wait.

Example prompts

• "Make a 4:5 picture of a paper boat on a rainy window sill, soft morning light."

• "Edit this photo: put her in a red jacket and make the background a snowy street."

• "Turn this product photo into a 5-second video with a slow camera push-in."

• "Make a 9:16 video of a vintage car cruising along the coast at golden hour, 10 seconds." (over 300 credits, so the assistant asks first)

• "Apply the Zombie Turn effect to my selfie."

• "Write and produce an upbeat synth-pop song about a summer road trip, female vocals."

• "Read this paragraph aloud in the Puck voice, newscaster style."

• "Remove the background from this logo, then upscale it to 4K."

• "What video models can you use, and how much does each cost?"

• "How many PopNow credits do I have left? Show my last five videos."

• "Find me a style for a pitch deck for our seed round." (find_skills, then use_skill; free)

A typical exchange:

You:        Make a 9:16 video of a vintage car cruising along the coast at golden hour, 10 seconds.
Assistant:  (calls create_video → "Needs confirmation … costs 370 credits (the user has 8450)")
            That's 370 credits from your PopNow balance. Want me to go ahead?
You:        Yes.
Assistant:  (calls create_video again with max_credits: 370 → job_id, then get_job)
            Here it is. It's saved in your PopNow library too.
You:        Now give me an upbeat synth-pop track for it.
Assistant:  (calls create_music → about 72 credits, under 300, starts at once)

MCP registry manifest

server.json describes the server in the MCP registry format (schema 2025-12-11): the remote Streamable HTTP endpoint with OAuth sign-in.

Troubleshooting

FAQ

How do I generate AI images and videos from Claude or ChatGPT?

Add https://api.popnow.ai/mcp as a custom connector (Claude) or custom MCP server (ChatGPT), sign in to PopNow, and ask in plain words. The assistant calls create_image, create_video or another PopNow tool and returns the result in the chat.

How do I generate a video from an image with an MCP server?

Give the assistant a photo (attach it in ChatGPT, or paste a public https link elsewhere) and describe the motion. It calls create_video with image_url, which switches to image-to-video. About 185 credits for 5 seconds on the default model.

Is there an MCP server for AI music and text-to-speech?

Yes. PopNow's create_music makes original songs or instrumentals (Suno v6) and create_speech reads text aloud in many voices, languages and delivery styles.

Does it cost extra to use PopNow in ChatGPT, Claude or Cursor?

No. Connecting is free, and each creation costs the same credits as on popnow.ai. Failed creations are refunded automatically, and anything over 300 credits waits for your confirmation.

Do I need to install anything?

No. PopNow is a hosted remote MCP server; you only add its URL. Clients that cannot connect to remote servers can use mcp-remote.

Can I use an API key instead of OAuth?

API keys belong to the developer server at https://api.popnow.ai/api/v1/mcp, billed in USD from your API balance rather than your website credits. See popnow-api-mcp.

Which AI models does the PopNow MCP server support?

Image, video and audio models including Kling, Seedance, Wan, Vidu, MiniMax, Nano Banana, GPT Image, Flux and Suno, plus one-tap video effects. Ask the assistant to call list_models for the current list with prices, then name a model in your request.

How do I find a ready-made style for a launch video, slide deck or landing page?

Ask the assistant, for example "find a style for our product launch video". It calls find_skills, shows you a few presets with preview links, and after you pick one calls use_skill to get its playbook and install link. Presets are free; only the generations they run cost credits. You can also browse the catalog at popnow.ai/skills.

Where do my creations go?

You get a link in the chat, and everything is saved in your PopNow library, just like creations made on the website.

Links

• Website: popnow.ai · Pricing · Use PopNow in ChatGPT and Claude · Skills

• Help: Use PopNow in ChatGPT and Claude · Use a skill in ChatGPT or Claude · Tutorials

• For developers: popnow-api-mcp, the API key MCP server billed in USD

Contributing

Issues and pull requests are welcome. See CONTRIBUTING.md. For security issues, see SECURITY.md.

License

MIT © PopNow