Skip to content

Vision

XPP Lab brings the self-contained XPPAUT experience back to more people. You write a model, simulate it, look at its phase plane, follow its bifurcations with AUTO, and understand it, all in one program that installs in a minute on any computer. XPPAUT has offered this for thirty years to the people who could get it running. XPP Lab makes it approachable for everyone else: students meeting differential equations for the first time, teachers, and researchers in neuroscience, biology, chemistry and engineering.

This page says where the project is going. The roadmap says how far it has got. The maintainer agent reads both when it proposes work.

Five themes

1. An app anyone can install and use

  • One download on Windows, macOS and Linux; no X server, no compiler, no terminal.
  • An interface that shows what each button does and why, with the equivalent command for every action, so the app teaches its own scripting.
  • XPPAUT users keep their .ode files, their keyboard shortcuts and their workflow.

2. Learn by doing: documentation and examples

  • A manual that teaches: tutorials that start from a question ("why does this neuron fire?") and end with a bifurcation diagram.
  • Many worked examples, from field to field, each runnable in one click.
  • Every feature documented where it is used: in the app, on the website and in Python.

3. The engine, extended

  • XPPAUT's numerics stay exactly as they are; new analyses sit on top of them.
  • More cases: larger models, more kinds of equations, and the XPPAUT features still missing.
  • Deeper analysis: sensitivity to parameters, Floquet multipliers, isochrons and phase reduction, codimension-two bifurcations, parameter fitting.

4. Beyond XPPAUT: control design

Most tools that analyse a system stop at describing it. XPP Lab should also let you change it:

  • linearise a model at an equilibrium, with chosen inputs;
  • see its eigenvalues in the complex plane, and whether it is controllable;
  • drag an eigenvalue to where you want it, get the feedback gain that puts it there, and simulate the controlled system;
  • root loci over a gain or a parameter, and optimal control (LQR) later.

5. Reach: open, citable, and ready for coding agents

  • Open source and easy to cite, with a paper, releases and a Python package on PyPI.
  • Ready for coding agents. Many researchers now work with assistants such as Claude Code. XPP Lab gives those agents their own tools: an MCP server, a skill and machine-readable results. An agent can then take an .ode model and run a whole analysis, while the researcher checks every step.

Rules that do not bend

  • The numbers do not change. Every example gives bit-identical output to unmodified XPPAUT 8.0.
  • Every XPPAUT model works unchanged.
  • Everything the app does can be done from a script, and the other way round where it makes sense.
  • Credit where it is due: XPPAUT is Bard Ermentrout's work, and AUTO is Eusebius Doedel's.

How work is chosen

The maintainer agent, The Limit Cyclist, proposes the next steps: a few at a time, each tied to a theme and a roadmap milestone, with why it matters and what done means. The project's maintainer approves or declines each proposal, and the agent then plans it, builds it, tests it and opens a pull request. Large features start with a short design note, which is agreed before any code is written. Small, finished steps are preferred to large, half-done ones.