Introducing the Fuul MCP: Run Your Incentive Program From Your AI Assistant

Managing an incentive program used to mean juggling tabs, exporting CSVs, and switching between dashboards just to get a clear picture, and a fresh round of clicks every time you wanted to change something. The Fuul MCP changes that. Once connected, your AI assistant becomes a full interface to your program: build triggers and incentives from scratch, set commission rates and volume tiers, monitor live performance, and review payouts, all in plain language.

‍

But the real shift is what happens after you launch. Your program stops being something you configure once and becomes something you steer in conversation: read what's working, change the structure, and see the result, without leaving the chat.

‍

One assistant, full program access

Stand it up

Setup is a one-time conversation. Tell your assistant what to track, who gets rewarded, and how. It pulls the available options and builds each piece with the right shape.

‍

Define what counts. Tell it which user action to track; it creates the trigger.

"Create a trigger for when a user makes a trade on our perps platform."

‍

Set up who gets rewarded. Link the trigger to a payout structure: end users, referrers, or both.

"Create an incentive that pays referrers 20% of the trading fees from the users they refer, and gives traders 5% of their own fees back."

‍

Reward your highest-volume referrers. Set commission tiers by referred volume. The more someone drives, the bigger their cut.

"Set commission tiers by referred volume over the last 30 days: 25% above $100K, 30% above $500K, 35% above $1M."

‍

Run it

The program is live. From here, the job is reading what's happening and adjusting, in the same conversation.

See what's happening. Ask a pointed question, get live data.‍

"Show me the top 10 affiliates by volume this month and how much each has earned in commissions."
‍

Let it spot the problem, and propose the fix. You don't have to know what's wrong, or how to fix it. Ask how the program is doing and the assistant reads the data, flags what's off, and lays out your options.
‍

You: "How are referrers performing this month?"

Assistant: "Referrer-driven volume is down about 30% over the last three weeks. A cluster of referrers are stuck just under the $500K tier with nothing pushing them across, and the long tail is flat".

Two ways to fix it:

1. Introduce competition. Add a weekly pool on top of commissions, say $25K split among referrers by their share of referred volume

2. Add intermediate tiers. Introduce steps in between so the next threshold is always within reach: 27% above $250K, 32% above $750K. Mid-tier referrers always have something to work toward.
‍

Or tell me what you have in mind and I'll set it up."

You: "Go with option 1, but make the pool $40K."

Assistant: "Done. Starting Monday, referrers split a $40K weekly pool proportionally to the volume they refer, on top of their usual commissions. Ask me next week and we'll see how it compares."

‍

In a dashboard, the data doesn't tell you it's slipping, where, or what to do about it. Here it's a conversation: the insight, the options, and the change, in one place.

‍

Handle relationships as they form. When a partnership closes off-platform, wire it up manually: assign the right referrer, set their commission tier, and lock it in for a fixed period.

‍

"Who referred wallet 0x1234? Set our LATAM partner as their referrer, and hold them in the 35% tier for 60 days even if their volume dips."

‍

Review payouts before approving. Pull pending payouts filtered by any criteria, review them in context, then approve from the app.

"Show me all payouts pending approval above $500 from affiliates active more than 30 days, with their total volume and earnings history."

‍

That's the Fuul MCP. The less friction there is to change something, the more you actually change it, and the better your program gets over time.
‍

Set it up in five minutes

Getting set up takes under five minutes. In Claude Code, run:
‍

/plugin marketplace add kuyen-labs/mcp_server

/plugin install fuul-mcp@fuul-mcp
‍

Then log in once from your terminal:

npx -y --package=@fuul/mcp-server@latest fuul-mcp login

‍

When prompted for the API URL, enter https://api.fuul.xyz/. Authenticate via browser and your session is stored automatically. Ask your assistant to find your project to confirm everything is connected.

‍

If you're using Claude Desktop instead of Claude Code, skip the commands above and add this to ~/Library/Application Support/Claude/claude_desktop_config.json:

‍

"mcpServers": {

  "fuul": {

    "command": "npx",

    "args": [

      "--package=@fuul/mcp-server@latest",

      "--yes",

      "--",

      "fuul-mcp-server"

    ],

    "env": {

      "FUUL_API_BASE_URL": "https://api.fuul.xyz"

    }

  }

}

‍

Then open a terminal and log in:

‍

npx -y --package=@fuul/mcp-server@latest fuul-mcp login

‍

Once authenticated, restart Claude Desktop and you're good to go.

‍

The Fuul MCP works with any MCP-compatible AI client. The setup above covers Claude Code and Claude Desktop.

‍

We'd love your feedback

Are you using the Fuul MCP to run your program? We'd love to hear how it's working for you. If there's a flow you want to automate or something you wish it could do, reach out at ecosystem@fuul.xyz. We build new capabilities based on what our clients need.

‍