Getting Started

Overview

The Ask Adverteyes MCP server lets your AI assistant query the Adverteyes Creative Intelligence database directly. Your assistant treats Ask Adverteyes as a knowledge source it can call upon to answer questions. Try it out! Ask a question about improving creative performance — by industry/brand, platform, audience, or your own portfolio — and it will query the server and return specific creative change actions to make, each with a lift figure showing the incremental probability that the change will help the ad drive sales.

Ask Adverteyes is on a standard MCP HTTP server and can be added via your assistant’s MCP settings: pass the API key in the header, and you’re done. The sections below walk you through it

Service authentication

The server supports two authentication methods, which can be used interchangeably:

API Key

You can sign up for credentials by submitting the Get API key form.

Pass the key with every request in the x-api-key header

OAuth 2.0 Bearer Token

The server also supports OAuth 2.0 Bearer tokens. This is the recommended method for clients such as Claude.ai custom connectors that have native OAuth support.

The server exposes standard OAuth 2.0 / OpenID Connect discovery endpoints so clients can auto-configure:

Authorization Code Grant flow

For clients that require manual OAuth configuration, use the Authorization Code Grant flow with the following parameters:

Parameter	Value
Client ID	`6ph5adqt8je2r0kjoq69pr7fd3`
Authorization endpoint	`https://preview-authentication.realeyesit.com/oauth2/authorize`
Token endpoint	`https://preview-authentication.realeyesit.com/oauth2/token`
Scopes	`openid email`

Clients that support OIDC discovery (e.g. Claude.ai connectors) will pick these values automatically (except of Client ID) from the /.well-known/ endpoints above — no manual entry needed.

Both x-api-key and Authorization: Bearer cannot be used simultaneously.

MCP Settings json example

Using API key (works with all clients):

{
    "mcpServers": {
        "adverteyes": {
            "type": "http",
            "url": "https://mcp.adverteyes.ai/mcp",
            "headers": {
                "x-api-key": "<YOUR_API_KEY>"
            }
        }
    }
}

Using OAuth (clients with native OAuth support, e.g. Claude Code, VS Code):

{
    "mcpServers": {
        "adverteyes": {
            "type": "http",
            "url": "https://mcp.adverteyes.ai/mcp",
            "oauth": {
                "clientId": "6ph5adqt8je2r0kjoq69pr7fd3",
            }
        }
    }
}

Example of setup and usage of MCP client in Claude products as skills

How to install the MCP-based skill and connect the Ask Adverteyes MCP server on every Claude surface.

The MCP-based skill has two parts that need separate setup: the skill itself (teaches Claude how to answer creative strategy questions) and the MCP connector. Both must be configured for the skill to work.

You can download the skill file here: adverteyes-creative-insights-mcp.skill (ZIP format).

Quick reference

Platform	Install skill	Connect MCP
Claude.ai (paid)	Customize → Skills → upload ZIP	Settings → Connectors → Add URL
Claude Code (terminal)	`unzip` into `~/.claude/skills/`	`claude mcp add-json adverteyes '{...}'`
Claude Desktop (free)	`unzip` into `~/.claude/skills/`	Edit `claude_desktop_config.json`
Cowork	Customize → Skills → upload ZIP	Customize → Connectors → Add
Claude API	Upload via Skills API	Pass `mcp_servers` in API call
Cursor	Place in `.claude/skills/`	Add to `.cursor/mcp.json`

1. Claude.ai (web / desktop app)

Install the skill

Go to Customize → Skills (or Settings → Features → Skills on some versions).
Make sure Code execution and file creation is enabled.
Click the “+” button, then “+ Create skill”.
Upload the adverteyes-creative-insights-mcp.skill file (it’s a ZIP — Claude accepts it directly).
Toggle the skill on.

For Team/Enterprise plans: an Owner must first enable Skills in Organization settings → Skills. They can also upload the skill there to provision it org-wide so every member gets it automatically.

Connect the MCP server

Go to Settings → Connectors (or the “+” button in chat → Connectors).
Click “Add” or “+ Add custom connector”.
Enter:
- Name: Adverteyes
- URL: https://mcp.adverteyes.ai/mcp
- Client ID: 6ph5adqt8je2r0kjoq69pr7fd3
Claude.ai supports OAuth — it will auto-discover the authorization endpoints and prompt you to log in.
Enable the connector in your conversation via the “+” button → Connectors → toggle Adverteyes on.

Use it

Just ask a question naturally in chat:

“When should the brand appear in my Facebook Stories ads?”

Claude will auto-detect the skill is relevant, call the MCP get_actions and list_asset_filters tools, and synthesize an answer with lift numbers.

2. Claude Code (terminal)

Install the skill

# Create the skill directory
mkdir -p ~/.claude/skills/adverteyes-creative-insights-mcp

# Unzip the .skill file into it
unzip adverteyes-creative-insights-mcp.skill -d ~/.claude/skills/adverteyes-creative-insights-mcp/

It’s live immediately — no restart needed.

Connect the MCP server

claude mcp add-json adverteyes '{
  "type": "http",
  "url": "https://mcp.adverteyes.ai/mcp",
  "headers": {
    "x-api-key": "YOUR_API_KEY"
  }
}'

Note : in case of OAuth use oauth property instead of headers property.

Verify

claude
# Then inside the session:
/adverteyes-creative-insights-mcp
# Or just ask naturally:
# "What music works best for 15-second mobile ads?"

3. Claude Desktop (free tier — manual config)

Install the skill

Same as Claude Code — place the files in the skills folder:

mkdir -p ~/.claude/skills/adverteyes-creative-insights-mcp

unzip adverteyes-creative-insights-mcp.skill -d ~/.claude/skills/adverteyes-creative-insights-mcp/

Connect the MCP server

Edit the Claude Desktop config file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json

Add this:

{
  "mcpServers": {
    "adverteyes": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.adverteyes.ai/mcp",
        "--header",
        "x-api-key: YOUR_API_KEY"
      ]
    }
  }
}

Fully quit and restart Claude Desktop. You should see the MCP indicator (hammer icon) in the bottom-right of the chat input.

Note: This method requires Node.js installed on your machine.

4. Cowork (desktop agent)

Install the skill

Go to Customize in the left sidebar → Skills → click “+” → upload the .skill ZIP file.

Connect the MCP server

Same as Claude.ai — Cowork shares the same connector infrastructure. Go to Customize → Connectors → Add → enter the URL https://mcp.adverteyes.ai/mcp and client id 6ph5adqt8je2r0kjoq69pr7fd3.

5. Claude API (programmatic)

Upload the skill

import anthropic
import base64

client = anthropic.Anthropic()

# Read the skill zip
with open("adverteyes-creative-insights-mcp.skill", "rb") as f:
    skill_data = base64.b64encode(f.read()).decode()

# Upload as custom skill
# See the Skills API Quickstart at:
# platform.claude.com/docs/en/agents-and-tools/agent-skills

Connect the MCP server and use the skill in a single call

response = client.beta.messages.create(
    model="claude-sonnet-4-20250514",
    max_tokens=2000,
    messages=[{
        "role": "user",
        "content": "What branding patterns drive the highest sales lift?"
    }],
    mcp_servers=[{
        "type": "url",
        "url": "https://mcp.adverteyes.ai/mcp",
        "name": "adverteyes",
        "authorization_token": "oauth_bearer_token"
    }],
    tools=[{
        "type": "mcp_toolset",
        "mcp_server_name": "adverteyes"
    }],
    betas=["mcp-client-2025-11-20"],
)

6. Cursor / VS Code / other MCP clients

Install the skill

These editors don’t have a native skill system, but you can place the SKILL.md in your project and reference it via Cursor Rules or similar:

mkdir -p .claude/skills/adverteyes-creative-insights-mcp

unzip adverteyes-creative-insights-mcp.skill -d .claude/skills/adverteyes-creative-insights-mcp/

Connect the MCP server

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "adverteyes": {
      "type": "http",
      "url": "https://mcp.adverteyes.ai/mcp",
      "headers": {
        "x-api-key": "YOUR_API_KEY"
      }
    }
  }
}

Note : in case of OAuth use oauth property instead of headers property.

Getting an API key

If you don’t have an Adverteyes API key yet, request one at: adverteyes.ai/resource/api-demo/#Form-API-Key

The demo key demo-preview-api-key-2024-05-17-limited works for testing with limited functionality.

Support for other OAuth flows

We support other OAuth flows (i.e machine to machine). Contact us for more details.

Powered by Adverteyes