Skip to content

DefendableCloud · Overview

DefendableCloud is the hosted proof vault for agentic work.

DefendableOS is the engine. DefendableCloud is the vault. The engine runs the rulebook; the vault stores and shares the proof.

The core primitive is the Defendable Run:

Inputs -> Evidence -> Execution -> Checks -> Verdict -> Approval -> Receipt

Every product surface exists to create, verify, approve, store, or share that Run. If a feature does not support that loop, it does not belong in v1.

SurfaceStatusPurpose
Marketing siteLiveExplains the cloud product and trust model.
Vault appLiveUser portal for projects, runs, evidence, checks, approvals, receipts, datasets, models, and account surfaces.
APILive foundationFastAPI backend for auth, projects, runs, receipts, ledger, datasets, models, cooks, incidents, membership, org/RBAC, and admin operations.
CLILaunch foundationTerminal workflow for the full Run lifecycle.
Dataset libraryLaunch foundationMembers-only dataset catalog, download grants, receipts, and quota controls.
Enterprise controlsLaunch foundationRuntime boot checks, RBAC, API keys, public receipt redaction, WAF guidance, backups, incident runbooks, and security docs.
LaneWhat it proves
Agent Work ReceiptsAn AI agent did the work it claimed, evidence was attached, checks ran, a verdict was minted, and a human approved or rejected it.
Dataset ReceiptsA dataset package or download grant was tied to a member, quota, object, hash, provenance boundary, and public/private receipt boundary.
Compute ReceiptsA cook, runner, model pin, or incident produced execution evidence and a hash-chained record.

Additional receipt kinds can ride the same per-org chain: cook receipts, model pin receipts, incident receipts, and restore-drill receipts.

Generate Receipt.

That is the action that closes the Run. Until a human approves, no run receipt should mint. Once approved, the receipt is hash-chained to the organization’s prior receipt and served as JSON, PDF, and a public share URL.

DefendableCloud is not a generic chatbot, LLM host, or project-management app. It is the audit and proof layer.

It should answer:

  • What work was requested?
  • What evidence was attached?
  • What rulebook checked the work?
  • What passed, warned, or failed?
  • Who approved it?
  • What receipt was minted?
  • Can another party verify the public proof without seeing private storage keys?
SurfaceURL
Marketinghttps://defendablecloud.com
Vault apphttps://app.defendablecloud.com or current Cloudflare Pages app host
APIhttps://api.defendablecloud.com
Docshttps://defendabledocs.com