Building a branded minigame used to mean briefing designers and developers, then waiting through build cycles. The Optimove MCP removes that step entirely: an AI assistant builds the game for you from a plain-language description.

The Optimove MCP lets an AI assistant — Claude, ChatGPT, Gemini, Cursor, Copilot, or any MCP-compatible client — generate, preview, and publish a playable minigame to your Minigames (formerly Adact) library, without you writing any code.

This guide will walk you through how to create a minigame with the Optimove MCP, from connecting your AI client to publishing the finished game to your Minigames library.

Note: You'll need an Optimove account with Minigames enabled and an MCP-compatible AI client. No extra setup is needed on the Minigames side — the game-creation tools appear automatically once your client is connected.

Step-by-Step: How to Create a Minigame with the Optimove MCP

Step 1: Connect the Optimove MCP Connector

Choose your client below to expand its setup steps.

Set up in Claude.ai

Open claude.ai and sign in.

Click your profile icon → Settings → Connectors.

Click Add custom connector.

Enter Name: Optimove and URL: https://mcp.optimove.net/mcp.

Click Add. Claude opens an Optimove sign-in page — sign in with your normal credentials and approve access.

In any chat, click + in the composer → Connectors → toggle Optimove on.

Set up in Claude Desktop

Open Claude Desktop.

Go to Settings → Connectors.

Click + → Add custom connector.

Enter Name: Optimove and URL: https://mcp.optimove.net/mcp.

Click Add and complete the Optimove sign-in flow.

In any chat, enable the Optimove connector from the + menu.

Set up in ChatGPT

Open chatgpt.com and go to Settings → Apps & Connectors.

Scroll to Advanced settings and toggle Developer mode on.

Return to Apps & Connectors and click Create.

Fill in Name = Optimove, a short description, MCP Server URL = https://mcp.optimove.net/mcp, Authentication = OAuth.

Click Create and complete the Optimove sign-in flow.

In any chat, click + → More → Optimove.

Set up in Cursor & other MCP clients

Add a remote MCP server entry pointing to https://mcp.optimove.net/mcp with OAuth authentication. The exact menu path varies by client — check your client's documentation for its connector or MCP-server settings.

Developers

The full technical reference, including the tool catalog and authentication detail, lives in the Optimove Developer Hub.

Set up in Microsoft Copilot

Microsoft Copilot works a little differently from the other clients. Instead of adding a connector directly, you build an agent in Microsoft Copilot Studio, add the Optimove MCP as a tool, and publish that agent to Microsoft 365 Copilot. Marketers then @-mention Optimove in the Copilot chat they already have open. Every user signs in as themselves, so their Optimove permissions and PII masking apply exactly as they do in the Optimove UI.

Verify against Microsoft's docs

Microsoft Copilot Studio changes often, so treat the steps below as general guidance and confirm them against Microsoft's current documentation (links at the end). These steps are for the new Copilot Studio experience — Classic works the same way, the clicks just differ.

Add the Optimove MCP to your agent. In Copilot Studio, open your agent's Tools page and select Add a tool → New tool → Model Context Protocol. Enter a server name and description, set the Server URL to https://mcp.optimove.net/mcp, and choose OAuth 2.0 → Dynamic discovery for authentication (the Optimove MCP supports dynamic client registration). Create the connection and add it to the agent.

Publish. Select the chevron next to Publish (top right) — not the Publish button itself. In the Publish agent dialog, select and configure Microsoft 365 Copilot (add Microsoft Teams too if you want both — they're separate channels). If the agent is missing a name, description, or instructions, the dialog says so — fix and continue. Select Publish and wait for it to finish. On success, the dialog offers to share your agent or add it to the organization catalog.

Roll it out. For just you or demos, use the share option — instant, no admin needed. For the whole org, choose add to the organization catalog, which sends it to your admin for review.

Admin approves (org rollout only). In admin.microsoft.com, go to Agents → All agents → Requests, pick the agent, and Publish. Admins can review capabilities, data access, and maker info first.

Use it in Copilot. Open Copilot (m365copilot.com, or Copilot inside Teams, Outlook, Word, Excel, or PowerPoint). Go to the Agent Store, find it under Built by your org, and Install. In any chat, type @ → pick Optimove → ask. The first tool call shows a Connect card — sign in with your Optimove user, pick your tenant, then return to the chat and ask again.

Each change needs a fresh publish

Any change to the agent needs a new publish before users see it, and each publish creates a new version.

Microsoft references

Step 2: Describe the Game You Want

Once connected, describe the game in plain language. For example:

"Build a 4×4 memory-match game with a 60-second timer, our brand's icons, and three prize tiers."

The AI generates a complete, self-contained HTML5 game — markup, styling, game logic, and the Optimove Gamify SDK all embedded — and creates it in Optimove on your behalf.

Protip: The more specific your description — grid size, timer, colors, prize tiers — the closer the first version will be to what you want.

Step 3: Preview and Iterate

Wait for the live preview URL — it arrives within about 60 seconds.
Open the URL and play the game yourself.
Ask the AI for changes — "make the timer 90 seconds," "swap in these colors" — and it regenerates the game and updates the same preview URL.

Please note: The preview URL stays the same as long as you keep iterating within about an hour of the last update. After that window, a new preview slot is created instead.

Step 4: Publish to Your Minigames Library

Once you're happy with the preview, ask the AI to save the game. This publishes it permanently to your Minigames library under the name you choose.

Please note: Publishing is not the same as activating. Saving only adds the game to your library — attaching it to a campaign and activating that campaign are still manual steps in Optimove. The AI can't schedule or send a campaign for you.

Screenshot 2026-07-15 at 12.42.30.png

How Minigame Creation Works

The MCP can build all three Minigames concepts: skill-based, luck-based, and knowledge-based games.

Note: Currently, only skill-based games send gameplay data back to Optimove. Luck-based and knowledge-based games can be built and published, but their data-back configuration isn't available yet — it will be added in an upcoming update.

Other Things the Optimove MCP Can Do

Search your game library: ask the AI to find existing games by title (fuzzy matching works) — it returns each match's game ID, title, and CDN URL, and can pull the full HTML source if you want to reuse or reference an existing game.
Go from idea to campaign: the same connector can look up customer segments, create Target Groups, and draft campaigns — so a single AI session can go from game concept to draft campaign end-to-end.

For technical implementation details, see the MCP Minigame Creation guide in the Optimove Developer Hub. For connector setup basics, see Optimove MCP Connector.