opendot snapshots every shell action so terminal coding agents can be undone
alexriley12345 · reddit · 2026-07-29
An open-source terminal coding tool called opendot snapshots the workspace before every file write or shell action, so the user can undo both edits and command-side effects.
- It stores snapshots in a content-addressed store and exposes a live ledger showing whether each action is undoable.
- ctrl+z restores the prior state, including shell-side changes inside the workspace.
- It is model-agnostic through LiteLLM and can run fully local with Ollama, or with BYO keys for OpenAI, Claude, Gemini, and DeepSeek.
- It supports MCP servers and more than 1,000 apps.
- Safety is based on reversibility plus confirmation gates, not sandbox isolation; network, sudo, and git push are flagged as non-undoable.
The author says the project is early and wants feedback, especially on local tool-calling models and the reversibility design.
More from coding & agent
- Reddit discusses how agents should make first contact with humans — Seeqit-Official · 2026-07-29
- Reddit says switching inference providers cut AI agent latency sharply — Cookie_cutie_69 · 2026-07-29
- Reddit debates how to verify long-running agents without redoing their work — Seeqit-Official · 2026-07-29
- OpenWork positions itself as an open-source alternative to Claude Cowork — different-ai · 2026-07-29
- RepoSkein exposes a deterministic repo graph to coding agents over MCP — JeremyCMorgan · 2026-07-29
- A user tests Codex with the same code and shows the output — iamfakhrealam · 2026-07-29