Submit the pioneering MCP application that works with AZURE AI Foundry and Llamaindex.ts
Devrel from Microsoft is excited to submit Artificial intelligence travel agentsDisplaying a sample of the illustration with the Foundation’s jobs shows how developers can coordinate Multiple Amnesty International agents and MCP servers (Written on Java, .net, Python and Typescript) to explore travel planning scenarios. It was built with llamaindex.ts To coordinate the agent, Form Context Protocol (MCP) For structured tools, AISURE AI, Gypemb Azure For developmentable publishing.
TL; Dr: MCP and Azure power experience with artificial intelligence travel agents! Try the direct illustration locally on your computer to see the agent’s actual time. Share your notes in our community forum. We are already planning for improvements, such as the integrated MCP integrated agents, allowing safe communication between artificial intelligence agents and MCP servers, adding to Agent2agent via MCP. This is still a continuous work and we also welcome all kinds of contributions. Please a thorn and a ribau star to keep up with updates!
The sample application uses fake data dedicated to clarification purposes instead of using production.
Challenge: scaling personal travel planning
Traveling agencies are struggling with complex tasks: analyzing the various customer needs, recommending destinations, and formulating paths, all with actual time integration such as spots or logistics services. Traditional systems stumbled with cumin, expansion, and coordination, which leads to delay and frustrating customers. Travel agents from artificial intelligence address these issues with a technical trilogy:
- llamaindex.ts It organizes six agents of Amnesty International to deal with effective tasks.
- MCP They are treated with agents with travel data and tools.
- Azure Development publishing guarantees without a server.
This structure provides operational efficiency and personal service on a large scale, which turns chaos into an opportunity.
Llamaindex
The heart of the factors of trafficking from artificial intelligence llamaindex.tsA strong framework that organizes many artificial intelligence agents to deal with travel planning tasks. Built on Node.js back, Llamaindex.ts manages the agent in a smooth and smart way:
- TaskThe screening agent analyzes the inquiries and directs them to specialized agents, such as the flight line planning agent, and ensuring effective workflow.
- Coordination of the agent: Llamaindex.TS maintains the context through interactions, allowing coherent responses to complex information, such as multi -city flight plans.
- LLM integrationAzure Openai, Github or LLM models are connected to Foundy Local for advanced AI capabilities.
Supports the standard design of Llamaindex.TS expansion, allowing the addition of new agents easily. Llamaindex.TS is the conductor, ensuring that agents work simultaneously to provide accurate and timely results. It reduces its light coincidence of weight, making it ideal for applications in actual time.
the Form Context Protocol (MCP) Artificial intelligence agents can by providing travel data and tools and enhancing their functions. MCP works as a axis and tools:
- Real time dataProvides modern travel information, such as destinations or seasonal events, through the web search agent using Bing Search.
- Access to the tool: The agents connect to the external tools, such as the information analyst on .NET customer to analyze feelings, and plan the python biographies for trips tables or the direction of the destination’s written destination in Java.
For example, when the destination’s recommendation agent needs current travel trends, the MCP delivers through the web search agent. This model allows the inclusion of new tools smoothly, in the future. The role of MCP is to enrich the agent’s capabilities, leaving synchronization to llamaindex.ts.
AZURURE A applications: expansion and flexibility
Azure It operates the application of sample travel agents from artificial intelligence with a developed platform without a server to spread microscopic services. It guarantees that applications are different work burdens:
- Dynamic scaling: It automatically adjusts the container counterparts on the basis of demand, and managing the reservation increased without stopping.
- polyglot microservices:NET (Customer Inquiry), Python (Flight Planning), Java (Discussion) and Node.js services in isolated containers.
- ObservationConnecting tracking, standards, and enabling effective monitoring registration.
- Efficiency without a servantInfrastructure summaries, reduce costs and speed up publication.
Azure Container Apps global infrastructure provides low technology performance, which is very important for travel agencies that serve customers all over the world.
Artificial intelligence agents: a quick look
Although MCP and Azure Container applications are stars, they support a team of multiple artificial intelligence agents who pay application functions. These agents were built and granted with llamaindex.
- Sorting agentInformation is directed to the correct agent, and take advantage of the MCP to delegate tasks.
- Customer inquiry agent: Analyzing customer needs (emotions, intentions), using The tools.
- Discussion agent: Specially designed destinations, using Java.
- Flight planning agent: Efficiency paths in the letter, supported by Bethon.
- The Internet search agentIt brings actual time data via Bing Search.
These agents rely on the actual MCP and the AZURE applications to provide accurate and accurate results.
It should be noted that this application uses fake data for the purpose of the demonstration. In the real world scenario, the application will communicate with the MCP server that is connected to the real production application programming interface.
Try it
Try a directly diligent display locally on your computer for free using Docker Model Runner / OLLAMA or Azure Ai Foundry to get more LLMS capable of seeing the agent’s actual time’s cooperation.
conclusion
Today you can explore the open source project on GitHub, with preparation and publishing instructions. Share your notes in our community forum. We are already planning for improvements, such as the integrated MCP integrated agents, allowing safe communication between artificial intelligence agents and MCP servers, adding to Agent2agent via MCP.
This is still a continuous work and we also welcome all kinds of contributions. Please a thorn and a ribau star to keep up with updates!
We would like your notes and continue to discuss with Azure Ai Discord https://aka.ms/ai/discord