Binance Agent OS Execution Clerk

Governed Execution.
Zero Uncontrolled Fills.

DeskFlow enforces capital boundaries, arbitrates live liquidity across Spot and Convert, requires explicit human authorization, and audits every fill to the basis point on an immutable ledger.

Deterministic Lifecycle Pipeline
Strict Mandate Mode Active
Stage 01
Mandate Shield
Extracts dynamic order parameters and enforces cash only limits under ten dollars.
Stage 02
Liquidity Router
Compares live Binance bookTicker depth against Convert pricing with fee isolation.
Stage 03
Dual Key Gate
Halts all execution until operator issues explicit syntax token for the parent order.
Stage 04
Blotter Ledger
Calculates basis point shortfall against decision mid and commits to immutable markdown.
Agent OS MCP Feed
BNBUSDT Mid: 765.5450
Spread: 0.0100 USDT
BTCUSDT: 87450.00
ETHUSDT: 2415.50
SOLUSDT: 148.80
Scoped Account: AGENTIC_SUB
Mandate Cap: $10.00 Max
BNBUSDT Mid: 765.5450
Spread: 0.0100 USDT
BTCUSDT: 87450.00
ETHUSDT: 2415.50
Quickstart Guide

First time user guide.

Three simple paths to evaluate, test, and trade safely with DeskFlow in under sixty seconds.

Path One · Web Assistant

Test in Your Browser

Explore the interactive product walkthrough below. Live pricing and execution run through Claude or Cursor connected to the official Binance Agent OS MCP.

Jump to Live Assistant
Path Two · AI Assistant Chat

Chat via Claude or Cursor

Connect DeskFlow globally to Claude Code or Cursor. Talk naturally to your AI assistant to evaluate trades, review live prices, and approve orders with your explicit confirmation token.

pip install git+https://github.com/Datwebguy/DeskFlow.git
claude mcp add --scope user deskflow -- deskflow mcp
Path Three · Terminal Command

Direct Protected Order

Run the MCP-backed clerk from any terminal. DeskFlow requires Agent OS market data and a returned execution result before it records a fill.

deskflow "Buy 5 USD BNB. Cash only."
How You Trade with Your AI Assistant
STEP 1
Natural Request
You prompt: "Buy 5 USD BNB. Cash only." DeskFlow intercepts before exchange routing.
STEP 2
Live Quote & Check
DeskFlow checks spending limit, pulls live Binance depth, compares Spot vs Convert, and issues ticket PO-001.
STEP 3
Explicit Approval
You reply: CONFIRM PO-001. Casual phrases like "yes" or "ok" are strictly rejected.
STEP 4
Settled Receipt
Agent OS returns the fill; DeskFlow isolates fees, computes shortfall, and records the verified result.
System Architecture

Engineered for absolute accountability.

Modern autonomous agents can produce erratic commands. DeskFlow introduces mathematical rigor, execution boundaries, and posttrade verification into the Binance Agent OS tool chain.

SAFEGUARD ONE

Immutable Mandate Boundary

DeskFlow loads governing rules directly from configuration. Regardless of prompt complexity, orders requesting margin, futures, or leverage multipliers are rejected before accessing any order routing endpoint. Parent order capital is strictly capped at ten dollars notional value.

Permitted Venues Convert · Spot
Prohibited Venues Futures · Margin · Perps
Capital Cap $10.00 Notional Max
SAFEGUARD TWO

Scoped Subaccount

All writes occur strictly within the dedicated subaccount. Main account balances, external transfer capabilities, and withdrawal methods are denied by protocol design.

SAFEGUARD THREE

Dual Venue Arbitrage

Real time evaluation compares Spot order book depth against Convert pricing. Guarantees true market liquidity selection with zero synthetic or fabricated quotes.

SAFEGUARD FOUR

Implementation Shortfall Audit

Following fill confirmation, DeskFlow calculates execution shortfall in basis points against original decision mid pricing, while keeping exchange fees isolated to ensure clean slippage measurement.

Shortfall BPS = ((Execution Price minus Decision Mid) / Decision Mid) * 10000
Live Agent Interaction

Chat with the clerk. Governed in real time.

This browser page is a product preview. For live Binance market data and execution, connect Claude or Cursor to DeskFlow and the official Binance Agent OS MCP.

D
DeskFlow Agent Clerk
Agent OS Active · Scoped to AGENTIC_SUB
Active Mandate Ceiling: $10.00 USD
DeskFlow Clerk · Agent OS
Ready to govern your parent orders. Live Spot/Convert data and execution are available through the connected Agent OS MCP workflow.
Integration Protocols

Model Context Protocol tool interfaces.

Connect DeskFlow directly to Claude Code, Cursor, Codex, or any client using standard input and output streams.

# Step 1: Install DeskFlow globally via pip
pip install git+https://github.com/Datwebguy/DeskFlow.git

# Step 2: Connect official Binance Agent OS gateway
claude mcp add --scope user binance-mcp-server --transport http https://agent.binance.com/mcp/agentic

# Step 3: Attach DeskFlow execution clerk globally
claude mcp add --scope user deskflow -- deskflow mcp

# Step 4: Chat naturally with your assistant
Operator: "Buy 5 USD BNB. Cash only."
DeskFlow: Emits PO-001 ticket with decision mid and touch ask.
Operator: "CONFIRM PO-001"
DeskFlow: Commits execution, isolates fee, and returns glowing receipt card.
# Step 1: Install DeskFlow globally
pip install git+https://github.com/Datwebguy/DeskFlow.git

# Step 2: Execute dynamic order prompt from any folder
deskflow "Buy 3 USD SOLANA. Cash only."

# Clerk evaluates live Binance bookTicker and asks for authorization
[MANDATE CHECK PASSED]
Parent Order ID : PO-001
Side            : BUY
Asset           : SOL
Notional USD    : $3.00
Decision Mid    : 104.0950 USDT
Selected Venue  : SPOT
AWAITING APPROVAL: Type 'CONFIRM PO-001' to commit.
# Run complete zero dependency verification suite from any folder
deskflow test

# Output verifies leverage rejection, size limit, live quotes, and receipt rendering
==================================================================
 DESKFLOW VERIFICATION SUITE · BINANCE AGENT OS
==================================================================
[STEP 1] Testing Rejection Gate (Leverage Denied) -> PASS
[STEP 2] Testing Rejection Gate (Size Exceeds Cap) -> PASS
[STEP 3] Testing Dynamic Cash Order -> PASS
[STEP 4] Testing Explicit Authorization -> PASS
{
  "tools": [
    {
      "name": "deskflow_price_order",
      "description": "Validates a parent order against the mandate and Agent OS Spot/Convert data",
      "parameters": {"order_text": "string", "market_data": "object"}
    },
    {
      "name": "deskflow_confirm_execution",
      "description": "Reconciles an authorized Agent OS execution result into the immutable blotter",
      "parameters": {"po_id": "string", "confirm_token": "string", "execution_result": "object"}
    },
    {
      "name": "deskflow_update_mandate",
      "description": "Updates the active session spending cap",
      "parameters": {"max_notional_usd": "number"}
    },
    {
      "name": "deskflow_inspect_blotter",
      "description": "Reads permanent blotter ledger with historical shortfall metrics",
      "parameters": {}
    },
    {
      "name": "deskflow_verify_mandate",
      "description": "Returns active governing rules and capital limits",
      "parameters": {}
    }
  ]
}
Operator Inquiries

Frequently addressed questions.

Understanding DeskFlow boundaries, subaccount security, and execution mechanics.

Why is human authorization mandatory?
Autonomous language models can generate unintended trades through prompt drift or hallucination. Requiring explicit syntax confirmation guarantees that capital is never committed without human intent.
How are exchange fees handled?
Trading fees are isolated in a dedicated column on the blotter. This prevents slippage calculations from conflating market impact with fixed fee structures.
Can DeskFlow access master account balances?
No. All operations are strictly bound to the AGENTIC_SUB subaccount. Master account funds and external withdrawal mechanisms remain completely unexposed.
What happens if market data feeds drop?
If public book ticker endpoints fail to return reliable depth, DeskFlow halts the execution pipeline immediately with a data rejection error instead of inventing artificial quotes.