Everygen MCP: How to Connect It to Claude, Cursor and ChatGPT

What is Everygen MCP?

Everygen MCP is an integration that lets a compatible AI assistant request media creation from the service at everygen.ai through Model Context Protocol. The endpoint displayed on its MCP landing page is https://everygen.ai/api/mcp, with account access handled through OAuth. It is intended to bring creation tools into a conversation or agent workflow, while the provider handles generation outside the assistant itself.

Do you need MCP to make a video?

Start with the result you want. If you have a prompt, a reference image and a clip to make, a browser workspace gives you a direct place to choose settings, submit the request and review the output. You do not need to configure a connection to a chat assistant first. On this independent site, you can explore our video generator and use the creation form in your browser.

MCP makes more sense when your creative work already happens in an assistant: you may want to develop a brief, request media, inspect task status and continue a larger workflow in one conversation. That is more than avoiding a tab switch. However, it introduces a connection to maintain, permissions to review and a second interface through which a request can be submitted.

The browser route is useful for an occasional creation or for comparing the settings visible in a form. MCP is useful when the assistant needs to coordinate several steps. Neither route removes generation time, model limitations or credit costs. Choose the interface that helps you control the request rather than assuming a connector changes the underlying model.

Everygen AI Studio at everygenai.io is independent of everygen.ai and Viewmax LLC. Accounts, subscriptions, assets and credits do not transfer between these services. A balance purchased elsewhere cannot be used here, and signing in here does not recover an account from another domain.

Three similar names: which service are you connecting?

Treat the complete domain as part of the product name. In particular, everygen.ai and everygen-ai.com differ in both punctuation and domain suffix: the second ends in .com, not .ai. Do not identify a service by its logo, the word Everygen, or a search-result title alone. For the history of the first brand, see the Viewmax to Everygen rename.

The table summarizes public documentation, not an authenticated connection test. Each provider describes its own integration; that does not establish a relationship between providers. The everygen-ai.com footer explicitly states that it is independent of everygen.ai and Viewmax LLC. We have not established the corporate relationship, if any, between viewmax.studio and everygen.ai.

For example, everygen-ai.com documents everygen_get_capabilities, while viewmax.studio highlights generate_video and get_task and also mentions list_video_models. These are clues for comparing documentation with a client’s discovered tools, not proof that a connection is authentic. Tool lists can change and unrelated services can use similar names.

Before approving access, compare the full endpoint, the authorization destination and the account being authorized with that provider’s current documentation. A documented legacy domain deserves verification rather than an automatic assumption that it is a different company. If something does not match, pause before supplying credentials.

Do not paste private keys into unknown configuration files, public prompts or shared repositories. This site provides an independent guide and browser workspace; it does not provide an MCP server or accept credentials for these other services.

Models and workflows described by everygen.ai

The MCP landing page lists eight model names: Seedance 2.5, Veo 3.1, MiniMax H3 Max, Grok Imagine, Kling 3.0 Pro, Gemini Omni Flash, Nano Banana Pro and GPT Image 2. This is the provider’s advertised roster, not a tested availability matrix or a list of the models enabled in this separate studio.

A model name alone does not specify accepted inputs, duration, resolution or a request’s cost. Check the options returned by the connected service before submitting. A prompt describing “4K” does not establish that a tool exports a 4K file. For this site’s own settings, read which model fits your project.

These are featured creative workflows, not four documented MCP function names. Their descriptions do not establish a fixed export resolution or duration for every request. Avoid copying a workflow label into a hand-written tool call unless the connected server actually exposes that name and schema.

Plan the story before you choose an interface. A useful brief identifies the subject, action, framing and intended ending. If the result needs several shots, decide which visual details must remain consistent and which shots can vary. Keep reference assets organized so you can distinguish an input from a generated result.

For an example of planning and reviewing clips in this browser workspace, see the short-form workflow we use. That article describes this site’s workflow, not a demonstration of the other provider’s MCP tools.

What do the seven media scopes allow?

The everygen.ai developer page groups media access into seven scopes. They describe permission boundaries: they are not function names, and granting one does not mean every model or operation is available. The page says clients can request a subset and that the service checks scope requirements when tools run.

Identity and session scopes, such as openid and offline_access, also appear in its OAuth metadata. Consequently, “seven scopes” here means the seven listed media and generation permissions, not the complete authorization vocabulary. The browser comparisons below describe related activities, not API compatibility.

Review the access requested for the workflow you intend to run. An image-only experiment should not be treated as a reason to grant unrelated permissions without understanding them. Likewise, an authorization error is not a reason to paste a token from another service into the configuration.

The checked developer page does not provide a complete named tool catalog with parameter schemas. After connecting, inspect the tools actually discovered by your client. Use their current input definitions rather than inventing names from the scope table.

How to approach setup in Claude, Cursor and ChatGPT

The following separates documented setup from verification still needed in your own account. We reviewed public pages without signing in, registering a client, purchasing credits or submitting a generation. The everygen.ai page names all three assistants, but its visible add-connector button is for Claude. That is not evidence of a completed connection in every client.

https://everygen.ai/api/mcp

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

Why do two Everygen pages show different endpoints?

Public documentation checked October 18, 2026: the everygen.ai MCP landing page shows https://everygen.ai/api/mcp, while its developer page links to https://www.viewmax.io/api/mcp. Its authorization-server metadata also identifies www.viewmax.io as the issuer. These are documented references associated with the rebranded service, not evidence of a fourth provider.

This observation does not establish that both addresses are interchangeable in every client. An OAuth connection involves more than the URL pasted into a form: the authorization issuer and protected resource must also match what the client expects. A redirect on a website homepage does not verify an MCP request or an OAuth exchange.

If a connection fails, compare the provider’s current endpoint and discovery metadata with your saved configuration. Check whether you selected the intended account, whether the authorization completed and whether the requested action has the necessary scope. Do not repeatedly swap domains or reuse a token across endpoints simply to bypass an error. These pages may change after the check date; use the source links below to review the latest information.

What does using Everygen MCP cost?

The everygen.ai MCP page does not publish a separate connection fee or a per-tool credit tariff. Its pricing page, checked October 18, 2026, displays the annual-billing rates below. These are monthly price equivalents billed annually; the credit allowances are monthly, not yearly totals.

That public pricing page lists paid plans and says plan credits refresh monthly without rollover. It does not list a free plan. This does not, by itself, prove that connecting MCP has a fee or that no account can receive a trial. Check your account’s current offer rather than reading an absent price as a billing rule.

Separate three questions: what your assistant subscription permits, what the media provider charges for its plan, and what the proposed generation consumes. We found no complete per-tool credit schedule in the checked MCP and developer pages. Marketing examples of how many outputs a plan can produce are not a fixed price for every model, duration or setting.

The similarly named everygen-ai.com publishes different terms: its MCP guide says connecting has no separate fee, its free plan includes 10 one-time credits, and generation uses an approved quote valid for ten minutes. Those statements belong to that domain and must not be copied into advice about everygen.ai.

Here at everygenai.io, eligible users can start with free credits and may choose paid memberships. Your current allowance and generation quote appear in the studio. Read how free access works here for eligibility and shared-credit rules. No credits from either external provider transfer into this workspace.

Before approving a creation, check the service name, selected model, output settings and quoted amount together. If the quote is missing or unclear, resolve that before sending another request. A successful connector installation alone does not answer any of those billing questions.

Frequently asked questions

If you need the provider’s MCP documentation

For current credentials, supported endpoints and billing terms, read the intended provider’s own documentation. These references support the comparisons above; they do not imply an affiliation with this site. The first three links concern everygen.ai. The other sources are labeled by service so their instructions are not confused.

• Everygen MCP landing page
• Everygen developer resources
• Everygen pricing
• Everygen OpenAPI transport specification
• Everygen authorization-server metadata
• EveryGen AI at everygen-ai.com: MCP documentation
• bytevirts connection package for everygen-ai.com
• ViewMax Studio MCP documentation
• Cursor MCP documentation
• OpenAI: add a custom MCP server