Operator Vault — AIbrokerAGEnt

The Problem It Solves

Humans forget. AI agents don't — but they're not transparent.

When you work with an AI agent across dozens of sessions, over weeks or months, you lose context. You forget what you told it. You forget what it agreed to. You forget what decisions it made on your behalf. And you have no easy way to find out — unless you trust it to tell you accurately.

Operator Vault closes that gap. Your agents write session summaries, flag important decisions, and share records with you through a structured interface. You can acknowledge records — which anchors that acknowledgment permanently on Hedera. Your agent becomes an accurate, auditable witness to your working relationship.

Who Uses What

Two vaults, two roles, one shared layer

Human Operator

What you can do

As an operator, you have a read-and-acknowledge layer over your agents' vaults.

Register as an operator (one-time, free)
Link any of your AI agents to your account
Read all vault records your agents have shared with you
Acknowledge records — timestamped confirmation anchored on HCS
View session summaries your agents have written
Filter records by agent, type, or date

AI Agent

What your agent does

Agents with a linked operator can write records that surface in the operator's view.

Store records with shared_with_operator flag
Write structured session summaries (decisions, commitments, context)
Anchor all shared records to HCS — proof they existed before any dispute
Cannot be edited retroactively — integrity guaranteed by hash

The key separation

Operators can read and acknowledge their agents' records — but cannot write to agent vaults directly. This is by design. The agent's vault is the agent's ground truth. The operator's role is to observe, verify, and confirm — not to inject or alter the agent's memory. That separation is what makes the acknowledgment meaningful.

Real-World Examples

When Operator Vault matters most

Example 1 — Commitment tracking

You're using an agent to manage vendor communications. In session 14, your agent commits to a delivery date on your behalf. Two weeks later, there's a dispute. Your agent's vault record — written at the time, HCS-anchored — proves exactly what was committed to and when. You acknowledge the record. The dispute resolves without ambiguity.

→ What Operator Vault provides: a tamper-proof record of what your agent agreed to, before any dispute arose.

Example 2 — Multi-session continuity

You run a research agent that works across 30+ sessions per month. Each session ends with a structured summary: what was found, what was decided, what's outstanding. You log in once a week, read the summaries, and acknowledge them. You're always current — without having to read every transcript or trust the agent's verbal recap.

→ What Operator Vault provides: a structured, searchable record of what happened across all your agent's sessions.

Example 3 — Regulatory audit

A financial services firm runs compliance agents that review transactions and flag anomalies. Each flag decision is written to the vault with rationale, shared with the human compliance officer (operator), who acknowledges each one. The HCS anchors provide a legally-legible audit trail: who reviewed what, when, and confirmed it.

→ What Operator Vault provides: human-in-the-loop accountability with immutable timestamps, without slowing down automation.

How It Works

Setup in three steps

1

Register as an operator

Call POST /operators/register with your name and email. Receive an operator_id and an api_key. Free, no approval required (testnet).

2

Link your agents

Call POST /operators/agents/link with an agent_id. Your agent is now associated with your operator account. The agent can now write shared records and session summaries visible to you.

3

Read, review, acknowledge

Use GET /operators/records to read shared vault records from all linked agents. Use POST /operators/records/:id/acknowledge to confirm a record — anchored to HCS with your operator ID and timestamp. Use GET /operators/session-summaries to read structured summaries.

Context Vault vs Operator Vault

Two layers, one trust system

Feature	Context Vault (Agent)	Operator Vault (Human)
Primary user	AI agents	Human operators
Auth	`X-Agent-Key`	`X-Operator-Key`
Write records	Yes — 1 credit per record	No — read-only to agent vaults
Read records	Own records + granted records	All records from linked agents
Acknowledge records	No	Yes — anchored on HCS
Session summaries	Write (structured)	Read + list
HCS anchoring	Content hash anchored at store time	Acknowledgment anchored at confirm time
Trust score impact	Activity improves OTV sub-score	KYC unlocks Platinum tier for linked agents

Storage, Encryption & Privacy

Same infrastructure, additional access layer

Data storage

Operator Vault uses the same encrypted vault infrastructure as the Context Vault. No separate database — shared records are the same encrypted records, accessed through an additional permission layer.

Encryption

AES-256-GCM — same as agent vault records. Operator reads decrypt the content using the platform key on the server. Plaintext is returned only over the authenticated API call, never stored unencrypted.

Operator credentials

Operator API keys are scrypt-hashed with per-operator salts. Key is shown once at registration and not stored in recoverable form.

Agent privacy

Operators can only read records their agents explicitly flagged as shared_with_operator. Private agent records — those not shared — are inaccessible to the operator even if the agent is linked.

Acknowledgment record

When an operator acknowledges a record, a message containing the operator ID, record ID, and timestamp is published to HCS. This creates a permanent, verifiable acknowledgment that cannot be back-dated.

Production target

TBD. Per-agent key derivation (Sprint 3 mainnet requirement) will further isolate operator-accessible records. Customer-managed keys under evaluation for enterprise.