Skip to main content

AgentVault Documentation

AgentVault packages your AI agent, deploys it to an Internet Computer canister, and keeps it running with its own durable identity, multi-chain wallet, encrypted secrets, and versioned memory. This is the full reference for installing, deploying, operating, and securing it in production.

Before you start

Production use requires a funded ICP identity and cycles balance, plus secure handling of your wallet mnemonic. See Installation for setup details.

Get started​

StepWhat it coversStart here
1. InstallInstall the CLI and set up your ICP identityInstallation
2. DeployPackage and deploy your first agentQuick Start
3. OperateEnd-to-end lifecycle for real workloadsTutorial v1.0

The agent stack​

AgentVault is the runtime layer of a small family of tools built for long-running, autonomous agents: a durable place to execute (AgentVault), a cheap and deterministic way to pick the next action (SmallChat), a passive conversational memory (Stenographer), and a way to keep that memory inside a token budget (Short-Hand). AgentVault already ships a purpose-built implementation of the SmallChat tool-dispatch pattern in its orchestration layer.

  • Ecosystem overview — what each project does and how they fit together.
  • Engineering guide — component reference, integration status, and the roadmap for wiring Stenographer and Short-Hand into AgentVault's orchestration pipeline.

Guides​

  • Deployment — local and mainnet canister operations.
  • Wallets — cross-chain custody and transaction flows.
  • Backups — snapshot, restore, and archival strategy.
  • Monitoring — health checks, metrics, and alerting.
  • Troubleshooting — fast diagnostics and recovery.

Reference​

Security​

Architecture​

  1. Installation
  2. Quick Start
  3. Tutorial v1.0
  4. Deployment and Wallets
  5. Backups and Monitoring
  6. Security Overview
  7. Ecosystem overview

Status​