About IVYX Studio
A local AI workspace an agent can operate, on a governed core.
IVYX Studio is a local AI workspace an agent can operate. Notebooks, datasets, models, a debugger, Docker, MLflow and the agents themselves are typed capabilities on one governed core, so a person and an agent do the same work through the same rule, and every call lands on one audit trail an evidence package can be built from.
One binary opens three windows: IVYX Studio, the workbench; the IVYX Agent window, a conversation over the same governed agent loop; and the IVYX Learn window, a tutor over notebook courses. The ivyx CLI ships inside it and puts the same gate in a terminal. The product is written for four audiences, the ML engineer, the data scientist, the LLM and agent developer and the student, and for a fifth reader, the operator or compliance function, who reads the trail rather than making the calls.
IVYX Studio is the first product built on the punica-editor substrate, a modular editor runtime whose capabilities ship as a marketplace of extensions. The substrate stays product-agnostic; IVYX is the AI/ML lab layer on top of it, and the first next-gen product from PunicaLab.
What we believe
Local-first
Your files, datasets and local models stay on your machine. Hosted services are opt-in, and a rule of deny egress: remote stops any governed call from leaving.
One workspace
Data, compute, models and code together, not a dozen disconnected tools.
Extensible by description
A capability comes from five origins: an extension, an ivy-node or agent, an instruction class, an MCP server, a skill. Describe what is missing and IVYX writes it as an ivy-node with a manifest you can govern.
Governed by default
AI-built and customer-built features pass the same policy gate before they run, a person is asked where a rule or an agent's write requires it, and every call is on the trail.