LedgerMind AI
runs finance actions with proof.

A trusted AI expense agent for receipts, reports, approvals, and payments. Every sensitive action is tied to a Terminal3-verifiable agent identity.

4 scoped finance agents.
TERMINAL3
100% sensitive actions logged.
AUDIT
0 raw payment secrets in prompts.
TEE
1-click human approval before pay.
POLICY

// PRODUCT

Everything for trusted
expense operations.

LedgerMind keeps chat, receipt scanning, categorization, reports, approvals, and audit logs separate, while Terminal3 binds sensitive actions to verifiable agent identity.

  ┌───────┐
  │ ◉ ◎ ○ │
  │ ◎ ◉ ◎ │
  │ ○ ◎ ◉ │
  └───────┘

Receipt Intelligence

Upload PDFs, images, and screenshots. The Receipt Agent extracts merchant, amount, date, and category.

  ┌─┐   ┌─┐
  │A├──►│B│
  └─┘   └┬┘
        ┌▼┐
        │C│
        └─┘

Approval Workflows

Payment requests stop at a human approval checkpoint before the Payment Agent can execute.

   ╔═══╗
   ║ ◈ ║
  ┌╨───╨┐
  │░░░░░│
  └─────┘

Agent Identity

Every agent action carries a Terminal3 DID, contract function, digest, and audit proof.

  │  ▄ 
  │▁ ▄ █ ▄
  │█ █ █ █
  └────────

Spend Analytics

Track category totals, daily trends, deductible expenses, and vendor concentration from one dashboard.

    .--.
   /    \
  | (  ) |
   \    /
    '--'

Multi-Agent Ledger

Receipt, Category, Report, and Payment agents stay permission-separated with explicit denials.

  GET /api
  ────────►
  ◄────────
  { data }

Exportable Reports

Generate monthly, business, tax, CSV, and JSON-ready summaries with evidence trails.

// TERMINAL3 FLOW

Three steps to
trusted action.

workflow.ts
1setEnvironment('testnet')
2await t3n.handshake()
3const did = await t3n.authenticate(auth)
4
Ready

// AGENT INFRASTRUCTURE

Built for accountable money movement.

LedgerMind separates every responsibility into a scoped agent so receipt parsing, categorization, reporting, and payments never share unnecessary permission.

ID

Identity Before Action

The server authenticates through Terminal3 before a protected workflow is recorded

OK

Approval Gates

Payment intents are prepared separately and wait for explicit human approval

[]

Immutable Evidence

Action digest, agent DID, contract function, and status are kept with the audit row

Receipt Agent

low
3 nodes

  ┌───┐
  │ ◉ │
  └─┬─┘
    │

Category Agent

medium
4 nodes

  ┌───┐
  │ ◉ │
  └─┬─┘
    │

Report Agent

medium
4 nodes

  ┌───┐
  │ ◉ │
  └─┬─┘
    │

Payment Agent

high
2 nodes

  ┌───┐
  │ ◉ │
  └─┬─┘
    │

Audit Writer

sealed
5 nodes

  ┌───┐
  │ ◉ │
  └─┬─┘
    │

Policy Engine

scoped
3 nodes

  ┌───┐
  │ ◉ │
  └─┬─┘
    │
4
Agents
TEE
Session
100%
Logged

// LIVE PROOF

Agent operations
with proof.

Terminal3 adapter online|--:--:--
0
Expenses tracked
Seed ledger plus uploads
0
Agent identities
Receipt, category, report, payment
0%
Protected actions
Audit row generated
0
Secret leaks
No raw keys in the browser
Live protected-action feed
nowPOST /api/chatcategory-agent200proof
1sPOST /api/receiptsreceipt-agent200digest
2sPOST /api/reportsreport-agent200signed
3sPOST /api/actions/protectedpayment-agent202approval

// INTEGRATION ECOSYSTEM

Connect finance data.
Expose only proof.

The app keeps AI, storage, deployment, and Terminal3 verification isolated behind server APIs so secrets never ship to the client.

  ┌─┐
  │#│
  └─┘

Terminal3

Agent Auth SDK

  ╔═╗
  ║<║
  ╚═╝

OpenAI

Assistant reasoning

  ┌$┐
  └─┘

MongoDB

Ledger storage

  [█]
  [█]

Vercel

App hosting

  ◈◈
  ◈◈

Cloudinary

Receipt media

  ≋≋
  ≋≋

CSV

Report export

  {M}
  ---

JSON

Audit export

  ▲
  ─

Render

Optional backend

Need a safer payment integration?

Add a real payment provider behind the Payment Agent. The agent can prepare a payment intent, but Terminal3-scoped approval remains the final gate before execution.

// Example: protected action
const proof = await ledgerMind.protect({
agent: "payment-agent",
action: "pay AWS bill"
});

// GOVERNANCE

AI finance with controls you can inspect.

LedgerMind is designed for high-stakes workflows where a model should never silently move money, hide actions, or see raw payment secrets.

  ╔═══╗
  ║ ◈ ║
  ╚═══╝

Terminal3 Session

The server opens an authenticated SDK session before protected actions

  ┌───┐
  │ ✓ │
  └───┘

Scoped Agents

Each agent declares allowed and denied responsibilities

  ╭───╮
  │ ★ │
  ╰───╯

Approval Required

High-risk payment actions are pending until a human approves

  [===]
  [===]

No Browser Secrets

OpenAI, Terminal3, and database keys stay server-side only

  ◉─◉─◉
  │ │ │

Action Digest

Each operation receives a SHA-256 digest and proof signature

  ▪ ▪ ▪
  ▪ ▪ ▪

Audit Logs

Every chat, upload, report, and agent action writes a trace

Terminal3-Centered Controls

Built around DID identity, scoped authorization, TEE execution concepts, and proof-first logging

SOC 2Type II
ISO 27001Certified
HIPAACompliant
GDPRCompliant
##

Production Safety Note

Payment execution is intentionally modeled as an approval-gated intent. Wire a payment provider only behind the Payment Agent after vendor policy, amount, and approver checks pass.

// SDK INTEGRATION

Built around the
Terminal3 SDK.

The Terminal3 adapter uses the real npm SDK for testnet environment setup, WASM crypto, ETH auth, session creation, usage reads, tenant helpers, and contract invocation paths.

TypeScript-first

Terminal3 SDK routes are typed and isolated in server-only modules.

TEE-ready contracts

The app exposes a clean path for tenant maps, secrets, contract registration, and invocation.

Audit by default

Every API action returns a proof envelope and writes an audit row.

Production deployable

Next API routes keep one Vercel deployment while still supporting a separate backend later.

1import { T3nClient, setEnvironment } from '@terminal3/t3n-sdk'
2
3setEnvironment('testnet')
4const client = new T3nClient({ wasmComponent, handlers })
$pnpm add @terminal3/t3n-sdk
added 1 package in 0.4s

Open the working LedgerMind app.

Try the dashboard, chat, receipt scanner, reports, activity logs, and agent monitor. Each workflow is backed by server APIs and Terminal3 proof envelopes.

Terminal3 testnet ready