Make video with Kling 3 without leaving your AI agent

The tool switch that breaks your flow

You are deep in a conversation with your agent, shaping copy or a storyboard, and then you need a short video. The usual move is to open a separate site, sign in, paste your idea again, wait, download the file, and carry it back. By the time you return, the thread of what you were doing has gone cold.

Figgshield removes that round trip. It is an MCP connector your agent talks to directly, so the request to make a video happens in the same place as the rest of your work. You describe the shot, the agent calls the tool, and the clip comes back as a link in the reply.

Nothing about your setup changes except that your agent gains a new skill. You keep writing prompts the way you always have.

One endpoint, any agent

There is a single address to connect: https://api.figgshield.ai/mcp. Add it once as an MCP server and your agent can see the tools for image and video from then on.

It works with Claude on desktop and in Claude Code, with Cursor and Windsurf, and with ChatGPT in Developer Mode on a paid plan. If your agent speaks MCP at all, it can reach Figgshield.

Because the connection lives at the agent level, every project you open can use it. You are not wiring this up per repository or per document.

Your first five-second clip

Once the server is connected, making a clip is a plain-language request. Try something concrete rather than vague.

The more you say about camera motion, subject, and light, the closer the result lands. "Slow push in," "handheld," and "locked off" all read differently to the model.

1. Tell your agent what to make: "Generate a five-second video, slow push in on a ceramic coffee cup steaming on a wooden table, morning light from the left."
2. The agent calls the Kling 3 tool with your description.
3. You get back a private link to the finished clip.

Writing a Kling 3 prompt that lands

Kling 3 responds to detail about motion and framing more than to adjectives about mood. Name the camera move, name the subject, name the light, and keep it to one clear action for five seconds.

A prompt like "aerial shot rising over a pine forest at dawn, mist between the trees, slow upward tilt" gives the model three things to hold: the vantage, the scene, and the movement. Compare that to "beautiful nature video," which leaves almost everything to chance.

If a clip misses, change one variable and ask again rather than rewriting the whole thing. Swap the camera move, or the time of day, and you learn what the model is reacting to. Each five-second try is 35 credits, so a few passes is a sensible budget.

When you want the same look faster, ask for the Kling 3 Turbo tier. It trades some polish for speed, which suits early drafts while you are still finding the shot.

Start from an image you already made

Figgshield also generates images with Nano Banana 2, and every image you create carries an id. Pass that id to the video tool and Kling 3 animates the still you already approved.

This is the fastest way to keep a look consistent. Generate the frame, get it right, then bring it to life without describing the whole scene a second time.

An image runs 8 credits at 1K or 2K and 16 credits at 4K, and it draws from the same balance as video, so you are not juggling two accounts.

What a clip actually costs

Image and video share one credit balance. A five-second video is 35 credits. There is no per-seat fee stacked on top and no markup on the models themselves.

Figgshield charges at cost. We pass through what the models cost and make our money only on credits people buy and never spend. That is the whole business, which is why there is no reason to push you toward a pricier tier.

You can start on the $5 Pro trial, which runs seven days and includes 150 credits, enough for roughly four video clips or a wider mix of images and video. There is no free tier. Cancel before day seven and you pay nothing more. See pricing or top up on the credits page.