Real-world uses of Axiomatic Operators' MCP servers

Each case study takes a real task from science or engineering, and follows a clear path from problem to solution. Each step has a prompt and output shown for ease-of-reproducibility.

Tips and tricks for doing this yourself with Axiomatic Operators

  1. Start with a good agentic IDE, like Claude Code or Cursor; or use a graphical approach like Dify.
  2. Run in agent mode (click toggle AI pane button). Choose claude-4-sonnet model for faster performance, claude-opus for more involved planning and GPT5 for deeper thinking but slower
  3. For jupyter notebooks in cursor, we find it best to use GPT5
  4. You can keep as markdown files your favorite prompts in ./cursor/prompts. These can then be easily used using the @
  5. Keep a rules file, which includes for example how the MCPs are used and when. It can also include the location of directories you want to use. An example is here
  6. Click settings button (top right) and then MCP to see the status of MCPs. If they appear in red, try a cursor restart or check if the API key is entered correctly. You can find the full mcp file if you click on "add new mcp server"
  7. Add the other MCPs you need – for example, zotero, semantic-scholar, arxiv, google-scholar, mathematica, mathematica-docs, etc.

Feedback Encouraged

What case studies would you like to see? Contact us with your ideas or find us on Discord.

Join our Discord Server