Skip to content

Your money, understood.

Your transactions.
Your agent.
Your say.

Let your agent answer questions about your spending. Choose the accounts it can read and how long it has access. Vantage takes care of the connection.

From question to answer
You

How much did I spend on Amex last week?

Budget assistant is asking

Review this request in Vantage
Purpose
Explain last week’s spending
You select
Amex ··1004
You allow
Transactions · 1 hour
You approved this account
Your agent’s answer

You spent $284.60 CAD on Amex last week.

Approved transactionsSnapshot refreshed 10 min ago

From a question to an answer

Ask a question.
Approve the access.

Your bank connection stays in Vantage. Your agent receives only the transactions you approve.

  1. 01

    Connect your accounts once

    Link through Plaid. Vantage collects transactions and refreshes the approved view in the background.

    Connect your first account
  2. 02

    Your agent asks. You decide.

    A named agent sends a review link. You choose the accounts and a time limit before it can read anything.

  3. 03

    Get the answer. Keep the control.

    Your agent reads through MCP or CLI. Open Vantage to see its access history or stop future reads.

Built around a separate read database

Your agent reads transactions.
Your bank connection stays private.

Bank connection and ingestion stay in Vantage’s operational systems. Agents read an approved snapshot from a separate PostgreSQL database instance.

Bank side · Vantage

Bank connection

Plaid access tokens
Operational database

Approved snapshot
Separate database instance

Transaction snapshot

Approved transaction data
Read-only for agents

MCP or CLI
Your agent

Answers with context

Only while your
permission is active

Plaid credentials stay in Vantage’s operational systems. Agents read approved snapshots from the separate database.

Read the architecture

Permission belongs to you

One human.
Multiple agents.
Separate permissions.

Give your budgeting assistant and expense helper their own accounts, time limits and purposes.

Manage agent access

Know which agent is asking

Review each agent’s name, installation ID and purpose before choosing which accounts to share.

Give access a deadline

Access defaults to one hour and is limited to twelve hours. When it expires, the agent needs your approval again.

See the reads. Stop the next one.

Review which accounts each agent read and when. Stop access for one agent or all agents anytime.

Works where you ask

Bring Vantage to
your agent’s workspace.

Use the MCP connector with Codex or Claude Code, or work directly with the CLI. The same account permissions apply across every interface.

Your next spending question

Give your agent the context.
Keep the permission.