Viewmax MCP for Claude — What It Connects, and What It Cannot Do Yet
The Viewmax MCP server lets Claude generate video, images and voiceovers directly from a chat. This page explains what the connector actually exposes, how to wire it up, and where its limits sit.
Independent site — not affiliated with Viewmax or viewmax.io. Or run Seedance, Veo and Kling →
What open video models are putting out right now
Clips published on X this month. Every one links back to the post it came from — press play, then go make your own above.
Slow-motion food drop shot, vertical — the kind of spot that normally needs a table-top crew
source post →Product bottle held in frame with citrus and a water drop, shot vertical for a feed
source post →A drink bottle caught in a burst of citrus and spray — a beverage-ad beat in 9:16
source post →Ingredients dropping into a blender, framed vertical for a recipe or product Short
source post →Cosmetic applicator dropping into a red set — a beauty-ad product beat, vertical
source post →Hands working dough on a board — human motion, not just a static product turn
source post →What the Viewmax MCP server is
MCP is the Model Context Protocol, a standard way for an AI assistant to call an outside service. The Viewmax MCP server is Viewmax's implementation of that standard, published at the endpoint https://www.viewmax.io/api/mcp.
Once connected, the Viewmax MCP integration means you stop switching applications. You describe a video in the chat window you already have open, the assistant calls Viewmax, and a finished clip comes back. The generation happens on Viewmax's side against your account and your credits.
Viewmax MCP is documented for Claude, ChatGPT and Cursor. Claude is the flagship path and the one most people are searching for, but the connector itself is not Claude-specific.
How to connect Viewmax MCP to Claude
- STEP 1
Open your Claude settings
In the Claude app or on claude.ai, go to Customize and then Connectors.
- STEP 2
Add a custom connector
Name it Viewmax and paste the endpoint URL https://www.viewmax.io/api/mcp into the connector configuration.
- STEP 3
Connect and sign in
Authenticate to your Viewmax account, and the Viewmax MCP tools become available to the assistant.
That is the entire setup. There is no local install, no API key pasted into a config file and no separate runtime to keep alive, because this is a remote MCP server rather than a local one.
What the Viewmax MCP integration actually calls
The tool exposed by Viewmax MCP for full video work is compose_video, and the name is informative. It is not a raw model call. It is a request for a composed piece: scenes, voiceover and captions assembled together.
A realistic Viewmax MCP request looks like asking for a 45-second vertical Short on a topic, with cinematic scenes, a voiceover and bold captions. The assistant plans the narrative beats, then hands the specification to Viewmax MCP, which does the generation and assembly and returns something downloadable.
Alongside video, the Viewmax MCP connector covers image generation and voiceover generation, so a chat request that needs a still or narration rather than a full clip is handled through the same connection.
What Viewmax MCP is genuinely good at
Removing the context switch
Scripting in chat and generating in a separate tab is a real friction cost, and Viewmax MCP removes it entirely.
Letting the assistant do the planning
Story beats, hooks and caption pacing are language work, which is what the assistant is already strong at. Viewmax MCP hands the rendering to a system built for it.
Iterating in conversation
Adjusting a brief and regenerating is faster in dialogue than in a form-based interface.
Composed output, not raw clips
Because compose_video returns an assembled piece, the result is closer to publishable than a bare model call would be.
Where the Viewmax MCP limits sit
Being clear about boundaries matters more than a feature list, and several are worth knowing before you build a workflow on Viewmax MCP.
It is not free. Every generation the Viewmax MCP connector triggers spends credits on your Viewmax account exactly as it would in the web interface, and a conversational workflow makes it unusually easy to burn through them, because asking again costs one sentence.
It is also not a control surface for fine editing. You are describing what you want in natural language; you are not framing a shot. For work needing precise per-shot control, the Viewmax MCP path is the wrong tool and the studio interface is the right one.
And it depends on the assistant behaving. The Viewmax MCP server does what it is asked, but the specification it receives is written by the model interpreting your prompt, which adds a translation step between your intent and the render.
Testing the output layer before wiring Viewmax MCP in
Every Viewmax MCP request ultimately produces generated video, and that output has to suit your subject matter for any of the convenience to matter.
The generator below runs open video models directly in your browser — free, no account, no card. It is not Viewmax MCP and does none of the composition: no voiceover, no captions, no scene assembly. It runs the underlying step, turning a still or a sentence into a moving clip.
That is the right thing to check first. If clip quality works for your content, a conversational path like Viewmax MCP is a genuine workflow improvement. If it does not, wiring up a connector only makes unsuitable output arrive faster.
Why Viewmax ships a remote connector, not a local one
Early integrations of the kind Viewmax ships were local: you installed a small server on your own machine, pasted an API key into a configuration file, and kept the process running. That approach still has its place, but it is a poor fit for video.
Video generation is heavy work that happens on rented hardware regardless of where the request originates. A local process would spend its life forwarding requests and waiting, which is a lot of setup friction for no benefit. Running the connector remotely means the Viewmax account you sign into is the account that gets billed, the capability updates when the service updates, and there is no version of the tooling on your laptop drifting out of date.
The trade is that you cannot use it offline and you are trusting the service with the requests you send. For generated marketing clips that trade is easy. For sensitive material it is worth thinking about, as it would be with the web interface too.
How a Viewmax chat workflow changes what you make
The interesting effect is not speed. It is that the cost of trying a variation collapses to a sentence, and that changes behaviour.
In the Viewmax web interface, each attempt has enough friction that people plan carefully and accept the third result. In conversation, asking for the same idea with a different tone or a different opening is almost free in effort, so you explore more. That is genuinely useful for finding a hook, and it is exactly how budgets evaporate, because effort stopped being the limiting factor while credits did not.
The habit worth building is deciding in advance how many attempts a piece deserves, then stopping. Viewmax over chat removes the natural friction that used to enforce that limit, and nothing replaces it except your own discipline.
Questions about Viewmax MCP
- What is the Viewmax MCP endpoint?
- https://www.viewmax.io/api/mcp, added as a custom connector.
- Does Viewmax MCP work with anything besides Claude?
- Yes — ChatGPT and Cursor are also documented.
- Does using Viewmax MCP cost credits?
- Yes. Generations bill to your Viewmax account normally.
- Can Viewmax MCP make images and voiceovers too?
- Yes, alongside composed video.
- Do I need to install anything locally?
- No. It is a remote MCP server configured by URL.
- Is a Viewmax account required?
- Yes. You authenticate during connector setup.
Check the output, then connect Viewmax
Viewmax MCP is a well-shaped integration: remote setup, composed output, and no context switching. Run a clip in the generator above with your own material first, and if the quality suits your content, the connector is worth wiring in.
Try the shot before you pay for it
Run it free in the browser — no account, no card. Then send the two or three hero shots to a premium engine.