Careem has opened its Model Context Protocol to developers and startups across the region, allowing AI agents to integrate with rides, food, groceries and payments. To mark the launch, the company announced a hackathon powered by Amazon Web Services, with the final taking place in Dubai on 14 November 2026.
At launch, Careem MCP offers tools across rides, food, groceries and payments. Builders can use it to create AI agents, assistants and integrations that help people plan a trip or build a shopping cart. Booking and payment then happen securely inside the Careem app, so people stay in control of every order. Documentation, a tool catalogue and an agent playbook are available at builders.careem.com.
To celebrate the launch, Careem is hosting Careem Builders ‘26: Hack to the future, powered by Amazon Web Services (AWS). The hackathon challenges builders to create something that makes everyday life simpler for people across the region, with Careem MCP as their starting point. The top 15 teams will compete live in Dubai on 14 November 2026.
Magnus Olsson, Co-founder and Chief AI Officer, Careem, said: “AI agents are changing how people get things done. With Careem MCP, we’re opening the services millions of people use every day to builders across the region, so they can create agents that help plan a trip or put together the weekly shop. The hackathon is our way of celebrating that, and we can’t wait to see what the region builds.”
The hackathon is open to developers and builders across the region, who can apply individually or in teams of up to three. Applications close on 31 October 2026 at 11:59 PM GST, with the top 15 teams selected to compete at an in-person final in Dubai on 14 November 2026.
Participants will have access to Careem MCP’s tools and documentation, alongside technical resources from AWS to support the development of their AI-powered solutions. These include free access to Kiro, AWS’s agentic AI development environment, optional prototype credits for Amazon Bedrock and Amazon Bedrock AgentCore, subject to approval, and training sessions led by AWS on building and deploying AI agents and connecting them to MCP servers.
