--- title: "Is there an MCP server for making AI ad creative?" description: "Yes: connect Wireflow's MCP server to Claude, Cursor, Codex or ChatGPT, ask for ad images or short video ads in plain words, and approve the credit cost." updated: 2026-10-23 --- Yes. Wireflow's MCP server, `https://www.wireflow.ai/api/mcp`, lets an AI agent such as Claude, Cursor, Codex or ChatGPT make ad images and short video ads with the image and video models on your plan. You ask in plain words, and the agent finds or drafts a workflow, quotes the credits when you ask, runs it on your account and hands back the files. A second connector, the Editor connector, puts a headline and captions on a video and renders the final cut. Before you start: - **Connect your agent.** Add the server and sign in with your Wireflow account. [Connect your agent](/docs/use-wireflow-from-your-agent) has the steps for each client. - **Have credits.** Runs use your credits. New accounts get a free video, more videos need a paid plan, and [pricing](https://www.wireflow.ai/pricing) lists each plan's price and monthly credits. - **Add the Editor connector if you want finished video cuts.** Its address is `https://www.wireflow.ai/api/mcp/editor`. See [Editor Connector](/docs/mcp/editor). ## How do I generate ad creative with my agent? Give the agent your product photo and a short brief, ask it to quote the credits, then say yes. A typical session goes like this: 1. **Share the product photo.** Paste a public `https` link to it. If the photo is only on your phone or computer, ask the agent for an upload link: `create_upload_link` makes a no-login page that puts the file on your board. 2. **Ask for the ad.** Name the product, the angle, the sizes and any exact words, as in the prompt below. 3. **Let it pick a starting point.** The agent can search templates with `list_templates`, copy one with `clone_template`, or draft a new workflow with `create_workflow_from_prompt`. A copy or a draft is saved as a private draft, and nothing runs. 4. **Check the cost.** `list_models` returns real model prices, so the agent can quote before it runs anything. The skills in the cost table below quote by default and wait for your yes. 5. **Run and collect.** `run_workflow` spends the credits, and `get_execution` returns the status and the file URLs. ```text Use Wireflow. Here is my product photo: https://example.com/mug.jpg Make three static ads for a ceramic travel mug, in 1:1, 4:5 and 9:16. Headline: "Hot coffee for the whole commute". Button: "Shop now". Quote the credits and wait for my yes before you run anything. ``` ## How do I make ad variations in 1:1, 4:5 and 9:16? Ask for all three sizes in one request. The [ai-ad-creative-generator skill](/docs/skills/ai-ad-creative-generator) writes distinct ad concepts and headline variants from your product photo and brief, then renders each concept in 1:1, 4:5 and 9:16 at once, with the headline and call to action spelled exactly. It quotes the credits, waits for your yes, and saves the set with test-ready file names to `./wireflow-outputs`. Without the skill, ask the agent to set the image model's aspect ratio for each size. Each model's node page lists the ratios it accepts. ## Can Claude make video ads for my product? Yes. Through the connector, Claude can turn a product photo into a short clip, with the photo as the first frame, or into a clip of an AI creator holding the product and speaking a line. New accounts get a free video, and more videos need a paid plan. Two skills package these jobs, and the cost table below lists both. One makes a 6-second product clip with a product-safe camera move, such as an orbit, a push-in, a light sweep or a turntable, and keeps the product unchanged from the first frame. The other makes a 6-second vertical talking clip with sound. It flags that the creator is AI so the ad can be labelled honestly, and it is not for real customer testimonials or for cloning a real person. ```text Use Wireflow. Turn this product photo into a 6 second vertical clip with a slow orbit around the mug: https://example.com/mug.jpg Keep the mug unchanged from the first frame. Quote the credits first. ``` ## How do I add a headline and captions to the video? Ask the agent to edit the timeline of a Video Editor node, look at still frames, then render. It reads the cut with `get_video_timeline` and changes it with `edit_video_timeline`, where titles, captions and motion live. `preview_video_frames` shows still frames at no credit cost. Then the Editor connector's `render_video` renders that one node from outputs that already exist, so no image or video model runs again. - **Start from a workflow that has a Video Editor node.** Ask the agent to find a template that already contains one with `list_templates`, copy it with `clone_template`, and confirm the Video Editor node with `get_workflow` before it spends a run. - **Captions need word timing.** The Speech to Text node (`audio:whisper`) returns word-level timestamps for captions. - **Music and voiceover come from audio nodes in the workflow,** such as `audio:elevenlabs_music`, `audio:lyria3_pro` and `audio:elevenlabs_tts_eleven_v4`. - **Render limits:** up to 180 seconds, 60fps and 1920x1080 in either orientation, with 2 renders a minute and 30 a day per account. - **Known limits:** an audio track's volume is a single number, with no fades or ducking, and there is no node yet for cutting or bleeping one word mid-line. ## How can I generate video ads programmatically through an API? Build the ad workflow once, then run it from your code with `POST /api/v1/workflows/{id}/run` and an API key, sending a new product photo URL and headline each time. [How do I run an AI workflow from my app](/docs/trigger-a-workflow-from-your-app) has the code in curl, JavaScript and Python. To let a client or a teammate run it with their own credits, publish it as an app, as in [How do I let others run my workflow via API](/docs/publish-a-workflow-as-an-api-endpoint). Plan for one limit: the Apps API cannot take video or audio inputs yet, so send images and text. ## What does an AI ad cost in credits? Each run costs the credits of the models in it. Skill prices come from the skills README, checked on 8 October 2026, and model prices from the [pricing page](https://www.wireflow.ai/pricing): | Job | Skill or node | Model | Credits | | --- | --- | --- | --- | | Three static ads, in 1:1, 4:5 and 9:16 | [ai-ad-creative-generator](/docs/skills/ai-ad-creative-generator) | Nano Banana 2 | 60 (3 images) | | 6-second product clip from a photo | [ecommerce-product-video](/docs/skills/ecommerce-product-video) | MiniMax Hailuo 03 Max Turbo | 48 | | 6-second vertical talking clip with sound | [ugc-ad-generator](/docs/skills/ugc-ad-generator) | Seedream 5 Lite + Veo 3.1 Lite | about 106 | | Product photo in new scenes | [product-photography](/docs/skills/product-photography) | Nano Banana Lite | 16 | | One image | Nano Banana 2 node | Nano Banana 2 | 20 per image | Resolution, length and quality change model prices, and every retry or extra variation with a skill needs a new yes. See also the [AI ad generator](https://www.wireflow.ai/ai-ad-generator). Last updated: 2026-10-23 --- Documentation index: fetch https://www.wireflow.ai/llms.txt for the full list of Wireflow docs.