Start here
Biosimulant is a workspace for investigating questions with your own agent. Keep the plans, Lab revisions, runs, results and evidence saved through its services together, then inspect the work, repeat a setup or continue it with another agent or teammate.
Underneath all of it is the Biosimulant runtime, an open source Python package. Studio, agents, the Hub and the Developer API all run the same labs on it. You can use any of them without ever installing it.
Pick your starting point
| If you want to… | Start with | Also useful |
|---|---|---|
| Start a recorded investigation and follow its saved work | Start a recorded investigation | Open Studio |
| Connect ChatGPT, Claude or Codex and give it a goal | Work with your agent | Which surface do I want? |
| Review previous work or continue with another assistant | Sessions and the research record | Teams |
| Browse public Labs and see which models they use and how they are connected | Browse the Hub | Labs, runs and files |
| Run a published lab from your own code or CI as a managed job, with the version and runtime recorded | Developer API quickstart | CLI, connect an agent |
| Know what a result actually proves before you report it | Model compatibility | Passports and evidence |
| Write a model in Python, or wrap one you already have in SBML, CellML or ONNX, and publish it as a versioned lab | Runtime quickstart | Create a BioModule |
Do I need to write code? No. Give your agent a goal and review the work saved in Studio. Use the Hub to discover published Labs when the question needs a runnable setup. The runtime and CLI are there for local development and automation.
Build a model. Run it anywhere.
One contract across supported modeling standards. Compose a model once, then run it locally for free or as a durable job in the cloud. It’s the same runtime either way.
| Step | What you do | Guide |
|---|---|---|
| Build | Subclass BioModule, declare typed ports, and implement execute(inputs, *, context). | Create a BioModule |
| Review in Studio | Inspect the saved setup, runs, results and evidence. | Studio |
| Run from an agent | Connect Claude Code or Codex over MCP and ask for what you need in plain language. | Codex and Claude |
| Run in the cloud | Submit the lab as a durable managed job through the Developer API. | Developer API quickstart |
| Run locally | Run your own modules next to version-pinned Hub labs. Free and MIT licensed. | Compose Hub labs locally |
The runtime doesn’t care whether your model is an SBML pathway, a CellML cardiac cell, a NeuroML neuron, an ONNX module or a few lines of Python. Each one implements BioModule, so any two can be wired together. Connect model_a.outputs.glucose to model_b.inputs.glucose and the runtime checks units, lines up time and delivers the signal.
Not sure which of these you’re in, or which one asks for a token? Which surface do I want? lays out all five side by side.
After your first quickstart, work through the section for whichever surface you’re using. Each one covers that surface on its own: Studio, agents, the Hub, the Developer API, model compatibility for what a result proves, and the Biosimulant runtime with its CLI underneath them all.