Introducing ConTree: Sandboxes That Branch Like Git
Nebius AI R&D is excited to introduce ConTree — secure, forkable code sandboxes as a service, tailored for research on AI agents and now available in early access.
Read article
Guides, deep dives, and updates from the ConTree team.
Nebius AI R&D is excited to introduce ConTree — secure, forkable code sandboxes as a service, tailored for research on AI agents and now available in early access.
Read article
The Contree SDK for Python is now publicly available. It wraps ConTree's execution platform into a small, typed Python client with async and sync interfaces, sessions, file handling, and a subprocess-compatible API.
Read article
Contree MCP is the Model Context Protocol server for ConTree. It gives AI agents isolated execution environments with persistent, immutable images and Git-like branching, so experiments are safe, repeatable, and cheap to explore.
Read article
Modern coding agents don’t just run code once. They explore, backtrack, compare, and verify. ConTree exists because that workflow needs a different execution model: results are artifacts, states can branch, and parallel hypotheses become practical.
Read article
ConTree is a sandbox execution platform designed for AI agents and developer tools that need safe, repeatable, and fast code execution. It combines VM-level isolation with Git-like branching so you can explore multiple paths in parallel and roll back instantly.
Read article