Credential management for AI agents

Your agents need secrets.
Keep them safe.

Zero-knowledge encrypted vault with native MCP interface. Agents request credentials through scoped permissions. Humans approve. Every access logged.

Start building View integration

Interactive demo

Try the full workflow in under 2 minutes

Create a vault, store a secret, then simulate an agent requesting access through policy controls.

Vault setup

Add secret

Agent access simulation

Agent: invoice-reconciler

vault_get("OPENAI_API_KEY")

Live vault state

No secrets stored yet.

mcp-config.json
{
  "mcpServers": {
    "agentvault": {
      "command": "npx",
      "args": ["@agentvault/mcp-server"],
      "env": {
        "VAULT_KEY": "lvk_..."
      }
    }
  }
}

Capabilities

Everything agents need.
Nothing they don't.

Zero-knowledge encryption

AES-256-GCM with Argon2id key derivation. Server never sees plaintext.

Scoped API keys

Read-only, time-limited, vault-scoped credentials with instant revocation.

Full audit trail

Every secret access logged with agent identity, timestamp, and source.

Human approval gates

Critical secrets require approval before retrieval.

Multi-agent support

Centralized policy management across autonomous workers.

Compliance ready

Audit exports in JSON/CSV with policy and access evidence.