🧪 Open Loyalty AI Labs: From challenge to prototype in weeks
Learn more
The Open Loyalty MCP server connects Claude, Cursor, and any MCP client straight to your loyalty engine. Describe what you want in plain language. The agent does the work. Your team keeps the judgment.
No endpoints to learn, no payloads to assemble. You describe the outcome, the agent maps it to real API calls, and you stay in control of what ships.
Ask in the words you’d use with a colleague. “Reward members who refer a friend.” No technical translation on your side.
It maps your request to real Open Loyalty calls like segments.create or campaigns.create across 22 domains, and shows its work.
The agent proposes the change with a structured result. Nothing touches your live program until a human signs off.
145 tools. 22 domains. One npx command.
Because Open Loyalty was API-first from day one, the whole platform is already addressable. The MCP server hands all of it to the agent, not a demo slice of it.
Real prompts from our own testing. The agent translates each into the right calls against your live program.
YouHow are our members distributed across tiers right now?
Agent + Open LoyaltyPulls tier analytics and shows how many members sit in each tier.
YouAdd 500 points to this member and confirm the new balance.
Agent + Open LoyaltyIssues the points through the API and reads back the updated wallet balance.
YouDraft an achievement for first-time referrers.
Agent + Open LoyaltyProposes the achievement rules and reward, ready to publish once a human signs off.
YouSet up double points this weekend for lapsed members.
Agent + Open LoyaltyBuilds the segment, drafts a time-boxed campaign, and holds it for your approval.
The agent operates inside the same boundaries your team does. You decide what it can reach, and you see everything it does.
It acts through your API token and store scope. It only reaches what you allow.
Changes are proposed and held for approval, not silently executed.
These are standard API calls, logged and traceable like any other request.
Server runs over stdio, the standard MCP transport. Nothing to host for local use.
We didn’t bolt an AI feature onto loyalty. We built an engine an agent can operate end to end: members, points, tiers, rewards, campaigns, segments, analytics. That starting point is why the MCP server reaches the whole platform, not a corner of it.
145 tools across 22 domains, including interactive views the agent can render — a live dashboard, member profile, or rewards catalog. If it lives in the Open Loyalty API, an agent can reach it.
Connect once with npx -y @open-loyalty/mcp-server and three values: your API URL, an API token, and a default store code. That scopes everything the agent can touch.
Steve proved AI can verify the real world for loyalty. The MCP server is the other half: giving an agent the controls to act on it. Both started here in AI Labs. The next one starts with your challenge.
Open source, MIT licensed, live on npm. Connect it to your AI client and start asking.