Deep Research Agent: Applications for Finance, Medicine, and ArXiv Research
Tip: Use the search box at the top of this page to find specific content.
Welcome to the The AI Alliance: Deep Research Agent for Applications project, which provides example applications using a powerful deep research agent called mcp-agentfrom LastMile AI. Several example applications are provided:
- Finance: Research the financials for a publicly-traded company.
- Medical: Research a disease or pharmaceutical.
- ArXiv: Research a topic by examining papers in ArXiv.org.
Other applications are planned…
All of them demonstrate the universality and flexibility of the deep research design pattern as implemented by the underlying mcp-agent tool kit.
Note: Do you have domain expertise, especially in finance, medical, legal, automation, industrial processes, etc.? Do you have AI agent expertise? Or, do you want to grow your expertise in these areas. Please join us! See our contributing page for details.
About
The applications leverage AI to perform automated research, analysis, and reporting. In fact, they are a thin veneer over a sophisticated agent framework, mcp-agent, a framework for creating AI agents with Model Context Protocol (MCP) integration, from LastMile AI, and application library code built on top of it.
The finance research application gathers data from multiple reliable financial sources to create structured investment reports with:
- Basic stock information and metrics
- Business overviews and revenue analysis
- Recent news and market events
- Financial performance summaries
- Risk and opportunity assessments
- Investor sentiment analysis
The medical research application currently uses a medical MCP server and web search, prioritizing known-reliable and freely-accessible sources, like PubMed.
The arxiv research application focuses its research on papers in ArXiv.org and also demonstrates the use of a tool called Docling for parsing documents like PDFs.
Other applications are planned. Possibilities include legal research, general science, industrial processes, including automation, etc.
The project README provides extensive information on running the applications, configuring them, and how to create new applications.
On our roadmap are plans to demonstrate techniques for testing AI-enabled applications, where model outputs are not deterministic, i.e., 100% predictable for a given set of inputs. Developers are accustomed to testing deterministic components and applications, but AI experts have the techniques developers can use to be confident their AI-enabled application works as designed. See Testing Generative AI Applications for more on this topic.
Try It!
The project README describes how to install the dependencies and run the applications. Once set up, the following make commands provide an easy way to get help on the commands and run them.
# For the default finance app:
make list-apps # List the currently supported applications.
make app-help-finance # Details on running the finance app.
make app-run-finance # Run the finance app with default arguments.
make -n app-run-finance # Show the command that would be run, but don't run it.
make app-medical-help # Details on running the medical app.
make app-medical-run # Run the medical app with default arguments.
make -n app-run-medical # Show the command that would be run, but don't run it.
make app-arxiv-help # Details on running the arxiv app.
make app-arxiv-run # Run the arxiv app with default arguments.
make -n app-run-arxiv # Show the command that would be run, but don't run it.
# For additional information:
make help # General help about the make targets.
NOTE:
The
app-run-*commands shown will invoke the inference service (OpenAI by default), incurring charges!