Skip to content

Let your AI agent
look after your saves.

Hoard runs no AI of its own. It ships a skill for the agent you already use: the agent reads your saves through the Hoard CLI, finds what is wrong and proposes the fix. Nothing changes until you say yes.

Get the CLI Read the skill

Works with any agent that can run commands, in the cloud or on local models

  • opencode open source
  • Claude Code
  • Codex open source
  • Gemini CLI open source
  • Cursor
  • Goose open source
  • GitHub Copilot
  • Ollama local
  • Cline open source
  • Windsurf
  • Roo Code open source
  • Zed open source
  • Amp
  • Kilo Code open source
  • Warp
  • OpenHands open source
  • Kiro
  • LM Studio local
  • Qwen Code open source
  • Junie
  • Trae
  • Antigravity
  • Qoder

Set up in one message.

The skill ships inside Hoard. Your agent installs it itself.

01

Install the CLI

Install it on the machine where your agent runs, then sign in once with hoard login.

Get the CLI
02

Paste this to your agent

It saves the skill wherever it loads skills from and reads it. That is the whole setup.

Run `hoard agents --skill` and save the output as a skill for yourself, in whichever directory you load skills from (for example ~/.claude/skills/hoard/SKILL.md). Then read it.
03

Ask in plain words

“Is everything backing up?” “Bring back yesterday's Factorio save.” “Why isn't this game syncing?” It reads first, shows you the plan and waits for your yes.

It asks before it touches anything.

Restores are rehearsed

Every restore runs as a dry run first and shows you, file by file, what would be overwritten. The real one waits until you confirm that save and that version.

Reading is free, changes are yours

Saves, versions, devices and diagnostics are read-only. Deleting, restoring and signing in always come back to you.

Always current

The skill lives inside Hoard and updates with it; your agent notices a new version and regenerates it. Hoard never writes into your agent's folders.