Biological foundation models move into Codex: natural-language prompts run perturbation experiments, with free compute for academics
Synopsis
Helical released a Model Context Protocol (MCP) integration for OpenAI's Codex that lets researchers call biological foundation models from the agentic tools they already use, and use or post-train those models on their own data through natural-language instructions. Using such models previously required researchers to configure infrastructure and build the workflows for an experiment. One model available at launch is PerturbGen, developed at the Wellcome Sanger Institute to predict how perturbations reshape cellular trajectories, and Helical with Nebius is offering free compute credits that academic researchers can request.
Interpretation
Biological foundation models can now be called from the agentic tools researchers already work in, without first standing up a separate runtime. Using these models previously required researchers to configure infrastructure and build the workflows needed to run an experiment. Helical's MCP integration brings those capabilities into OpenAI Codex, so researchers can use and post-train models on their own data through natural-language instructions.
Once connected, a concrete biological prediction can be made inside a conversational tool rather than a self-built pipeline. The Model Context Protocol is an interface convention that lets agentic tools call external models and capabilities, and Helical uses it to connect biological modelling to Codex. A joint quick-start example uses PerturbGen from Codex to predict how knocking out IL1B early in an LPS response alters inflammatory programs and cell states over time.
The release extends Helical's biological modelling beyond its existing pharmaceutical deployments into tools used by the wider research community. Helical adapts biological foundation models to proprietary data for pharmaceutical companies, for target identification and biomarker discovery. Helical and Nebius are providing free compute credits that academic researchers can request through a research-credit application form.
Perspective
The demonstrated scope is Helical's MCP integration for OpenAI Codex and one joint quick-start example that uses PerturbGen from Codex; the intended readers are researchers who want to use or post-train biological foundation models on their own data, and academic researchers can request free compute credits from Helical and Nebius through a research-credit application form.
The material reports no evaluation of prediction accuracy for the integration and does not describe plans beyond the models available at launch; the PerturbGen example is described only as predicting how knocking out IL1B early in an LPS response alters inflammatory programs and cell states over time, with no result reported.