Getting Started

Scenario MCP gives AI agents — Claude, Cursor, VS Code, or any MCP-compatible client — full access to the Scenario creative AI platform. 99 tools covering image, video, 3D, and audio generation across 750+ models (+ your own), with an intelligence layer that handles model selection, parameter discovery, and prompt enhancement automatically.

Prerequisites

Scenario account — Sign up at app.scenario.com
API credentials (API key auth) — app.scenario.com/team
MCP-compatible client — Claude Code, Cursor, VS Code, Windsurf, Claude Desktop, and more

Installation

Connect the Scenario MCP server to your preferred MCP-compatible client. OAuth is the recommended method — it handles authentication automatically on first use. By default the server lists a lean core of tools; the full catalog stays available on demand — see Missing a tool? if you expected more.

Multiple teams or projects

As a reminder, teams are shared workspaces for members, projects organize the assets and work inside that team. Please see Teams (also labeled as "Organizations") and Projects for more information.

If your Scenario account has access to multiple teams or projects, the Scenario MCP won't assume a default context. Instead, it instructs your AI agent to ask you which team and project to use.

If multiple teams or projects are available, your agent should ask which workspace to use.
Once selected, the agent can reuse that team and project context for later tool calls that need it.
If your client does not prompt automatically, ask your agent to list your teams and projects, then choose one explicitly.

Client-specific instructions

Claude Code

claude mcp add --transport http scenario https://mcp.scenario.com/mcp

OAuth authentication triggers automatically. Authenticate manually via /mcp if needed.

Claude Desktop

On Free, Pro or Max, open Customize → Connectors → Add → Add custom connector and paste:

https://mcp.scenario.com/mcp

Complete OAuth in the browser. Custom Connectors added on claude.ai sync to Claude Desktop too, so if you already added Scenario there, it is already connected here. On Team or Enterprise, your organization adds it once (see Claude.ai below); click Connect on it in Customize → Connectors.

If you cannot add a custom connector, use the mcp-remote bridge instead. Open Settings → Developer → Edit Config and add this entry to your claude_desktop_config.json file:

{
  "mcpServers": {
    "scenario": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.scenario.com/mcp"
      ]
    }
  }
}

Quit and restart Claude Desktop. OAuth opens in the browser on first use.

A URL-only entry ({ "url": "..." }) in claude_desktop_config.json is not supported by Claude Desktop; the app only loads stdio servers (command + args) from that file.

Claude.ai

On Free, Pro or Max, open claude.ai → Customize → Connectors → Add → Add custom connector, and paste:

https://mcp.scenario.com/mcp

Complete OAuth in the browser. The Free plan allows one custom connector.

On Team or Enterprise, the organization adds Scenario once: an owner (on Enterprise, also anyone whose custom role includes Manage access to Libraries) opens Organization settings → Connectors → Add → Custom → Web and pastes the URL. Each member then clicks Connect on it in Customize → Connectors and signs in with their own Scenario account; anyone else who tries to add it sees Ask an organization owner. Keep it on OAuth: a fixed API-key header on an organization connector gives every member that key's access. Until it is added, Claude Code connects without it.

Cursor

Cursor 1.0+ supports native OAuth and Streamable HTTP. Open Settings → MCP and add:

{
  "mcpServers": {
    "scenario": {
      "url": "https://mcp.scenario.com/mcp"
    }
  }
}

OAuth triggers on first use.

VS Code

Press Cmd+Shift+P, type MCP: Add Server, select HTTP, and enter the URL. Or add to your settings:

{
  "mcpServers": {
    "scenario": {
      "type": "http",
      "url": "https://mcp.scenario.com/mcp"
    }
  }
}

Activate via MCP: List Servers if not auto-enabled.

Windsurf

Open Cascade (Cmd+L) → Configure MCP and add:

{
  "mcpServers": {
    "scenario": {
      "url": "https://mcp.scenario.com/mcp"
    }
  }
}

OAuth triggers on first use.

Zed

Open settings (Cmd+,) and add:

{
  "context_servers": {
    "scenario": {
      "url": "https://mcp.scenario.com/mcp"
    }
  }
}

Zed prompts the OAuth flow on first use when no Authorization header is configured.

Codex

codex mcp add scenario --url https://mcp.scenario.com/mcp

OAuth triggers on first use.

Gemini CLI

Edit ~/.gemini/settings.json:

{
  "mcpServers": {
    "scenario": {
      "url": "https://mcp.scenario.com/mcp"
    }
  }
}

Run /mcp auth scenario to authenticate.

Amp

amp mcp add scenario https://mcp.scenario.com/mcp

OAuth authentication via Dynamic Client Registration.

OpenCode

Edit ~/.config/opencode/opencode.json:

{
  "mcp": {
    "scenario": {
      "type": "remote",
      "url": "https://mcp.scenario.com/mcp",
      "oauth": {}
    }
  }
}

Run opencode mcp auth scenario to authenticate.

Warp

Open Settings → AI → Manage MCP servers → + Add → JSON. Paste:

{
  "scenario": {
    "url": "https://mcp.scenario.com/mcp"
  }
}

OAuth triggers on first use.

Vercel v0

In v0.app open Account Settings → Integrations. Under MCP Connections click Add MCP and paste:

https://mcp.scenario.com/mcp

Complete OAuth in the browser.

Other Clients

Any MCP client that supports OAuth and Streamable HTTP can connect directly to:

https://mcp.scenario.com/mcp

For clients that only support stdio, use npx mcp-remote@latest https://mcp.scenario.com/mcp as a bridge.

Using an API Key

If you prefer API key authentication instead of OAuth:

Get your API key and secret from app.scenario.com
Base64-encode your credentials: echo -n 'KEY:SECRET' | base64
Add the Authorization header to your config

Claude Code

claude mcp add --transport http scenario \
  https://mcp.scenario.com/mcp \
  --header "Authorization: Basic $(echo -n 'KEY:SECRET' | base64)"

Codex

codex mcp add has no header option, so add the server to ~/.codex/config.toml:

[mcp_servers.scenario]
url = "https://mcp.scenario.com/mcp"
http_headers = { "Authorization" = "Basic <YOUR_BASE64_ENCODED_KEY:SECRET>" }

URL-config clients (Cursor, VS Code, Windsurf, Gemini CLI, OpenCode)

{
  "mcpServers": {
    "scenario": {
      "url": "https://mcp.scenario.com/mcp",
      "headers": {
        "Authorization": "Basic <YOUR_BASE64_ENCODED_KEY:SECRET>"
      }
    }
  }
}

Zed

{
  "context_servers": {
    "scenario": {
      "url": "https://mcp.scenario.com/mcp",
      "headers": {
        "Authorization": "Basic <YOUR_BASE64_ENCODED_KEY:SECRET>"
      }
    }
  }
}

Warp

{
  "scenario": {
    "url": "https://mcp.scenario.com/mcp",
    "headers": {
      "Authorization": "Basic <YOUR_BASE64_ENCODED_KEY:SECRET>"
    }
  }
}

Claude Desktop

Claude Desktop only loads stdio servers from claude_desktop_config.json, so the API key path uses mcp-remote as a bridge:

{
  "mcpServers": {
    "scenario": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.scenario.com/mcp",
        "--header",
        "Authorization: Basic <YOUR_BASE64_ENCODED_KEY:SECRET>"
      ]
    }
  }
}

Working in Blender?

The Scenario for Blender extension ships its own local MCP server.

It runs inside Blender on a loopback address with a per-session token and gives your agent scene tools (scene summary, selection, screenshots, quick renders, camera paths) plus generation and import straight into the open scene. Add it to your client next to this server: this server covers the Scenario platform, the extension covers what happens inside Blender. Client snippets and the tool list are in the scenario-labs/blender-plugin repository (GPL-3.0-or-later).

Agent Skills (optional)

The public Scenario Agent Skills teach your agent the call sequences behind common workflows. Install them locally:

npx skills add scenario-labs/skills

Without a local install, a connected agent can still read each skill from the server as an MCP resource at skill://scenario/<name>/SKILL.md.

Test the Connection

Try these prompts to verify everything works:

"List my Scenario teams"
"Show available models for image generation"
"Generate an image of a sunset over mountains"

Having trouble? See troubleshooting.

First Steps

Once connected, try these prompts in your AI Agent:

"Generate a pixel art knight with a red cape"

"What's the best model for YouTube thumbnails?"

"Show me my recent assets"

See what's new: MCP Server Changelog