AgentVault

Safe Obsidian vault processing from upload to verified output.

Safe AI cleanup for Obsidian vaults

Deterministic fixes first, guarded AI second. Review every change and download a cleaned vault.

Built for trust

  • Structure preserved: frontmatter, wikilinks, embeds, and code fences.
  • Clear review path with per-note report rows and diff viewer.
  • Guarded processing with deterministic fallback when checks fail.

How It Works

Three steps from raw vault to verified output.

Step 1

Upload

Upload your Obsidian vault zip to start a processing job.

Step 2

Repair & validate

Deterministic repair runs first, then guarded AI rewrite and structure checks.

Step 3

Review diffs + download

Inspect per-note diffs and report details, then download the cleaned vault zip.

Why It’s Safe

Core safeguards for predictable vault cleanup.

  • Preserves frontmatter, links, embeds, and code blocks
  • Deterministic cleaning before LLM
  • Quality guards with automatic fallback
  • Per-note report and diff viewer

What You Get

Output bundle from each completed job.

Cleaned vault zip

Preview.md

Report JSON

index.md + entities.md

Privacy: files are processed on our server for the active job, we do not train on your data, and you should avoid uploading vaults with secrets.

Created By

Created by Haseeb Raza

GitHubLinkedInX