Tilde.run is an agent sandbox built around a transactional, versioned filesystem. Instead of letting an agent make unreviewable changes, it tracks filesystem state so you can see what the agent did and roll it back. Because the sandbox is transactional, a run that goes wrong can be reverted to the state before it started. It is aimed at developers running coding agents who want a safety net underneath file edits rather than trusting them blindly.
Be the first to rate Tilde.run. Your feedback helps other users decide.
Give an agent an isolated place to work
Revert a run to the state before it started
See exactly which files an agent touched
Similar tools in the same category — compare side-by-side
Zot is a coding agent harness delivered as a single binary. There is no runtime to install, no Docke...
Boxes.dev gives every coding agent its own cloud computer. You bring your Codex or Claude subscripti...
Shoehorn makes a language model fit the memory you actually have. Preset quantizations ignore your h...
Guides and comparisons about code tools
Eleven new tools for running coding agents: sandboxes, agent version control, local runtimes and quantizers launched between April and October 2026.
AI code tools for web developers in 2026: frontend scaffolding, framework-aware suggestions, and review helpers that respect your design system.
AI pair programming tools compared: suggestions, chat, and autonomous edits, and how to pick one that fits your editor and review style.
Ten October 2026 agent infrastructure launches: local inference engines, sandboxes that report their isolation boundary, agent memory and MCP testing.
Eight local inference projects that fit big models into small memory: a 26B model in 2 GB, MoE weight streaming, a console as a server and sub-1-bit compression
