Programmatic Trading
AI Agent Dashboards

Five enterprise dashboard views for the 10 Programmatic Trading sub-agents. Each dashboard presents data from 20 MySQL tables in a format optimized for different audiences and use cases.

10
Trading Agents
20
Data Tables
5
Dashboard Views
100+
Data Columns
These dashboards display realistic demo data for demonstration. In production, the agents connect to your real ad tech data via MCP services or CSV import.
Deployment Options
The entire platform is available as a self-hosted solution or managed service
Self-Hosted
RECOMMENDED FOR ENTERPRISE
Deploy the entire platform on your own infrastructure. Your data never leaves your environment. Bring your own OpenAI API key. Full source code delivered.
Complete source code (Python agents, PHP dashboards, MCP services)
Data stays on your servers — no external data transfer
MCP connectors for your DSP, SSP, DMP, and ad server APIs
Custom integration and onboarding support available
Managed Platform
FOR AGENCIES & TRADING DESKS
We host and operate the platform for you. Upload data or connect your platforms via secure MCP services. No infrastructure management needed.
Fully managed — no DevOps required on your side
Secure data upload or API-based MCP integration
Dashboard access with your own branded login
Automatic updates and new agent releases
Both options include MCP services with connectors for your ad tech stack. The 10 data dictionaries in data_dictionaries/ define the schema contract between each agent and your data sources.
V1
Command Center
Real-Time Monitoring View
Monitors all 10 trading agents simultaneously with live status indicators, row counts, and data freshness checks. Provides collapsible detail sections for each agent with table purpose explanations and column guides.
Launch Command Center
V2
Executive Dashboard
Executive Summary View
Presents all 10 agent outputs in a single scrollable page for executive review. Each agent section includes key metric highlights, table purpose explanations, and column guides. Optimized for printing and PDF export.
Launch Executive View
V3
Analytics Explorer
Deep-Dive Analysis View
Enables deep-dive analysis of each agent's output with statistical summaries (avg, min, max, stddev) for numeric columns. Tab-based navigation lets analysts focus on one agent at a time with table purpose explanations and column guides.
Launch Analytics Explorer
V4
Operations Monitor
Operational Health View
Monitors agent operational health with row counts, data freshness, and system status for all 10 agents. Tabbed interface lets operators drill into individual agent tables with purpose descriptions and column guides.
Launch Ops Monitor
V5
Enterprise Report
Client Deliverable View
Structured as a formal report with numbered sections, table of contents, executive summary, glossary, and full data dictionary appendix. Each agent section includes table purpose explanations and column guides. Designed for client delivery as PDF.
Launch Report View

Platform Architecture

10 Input CSVs
Domain intelligence from page types (/api, /partners, /docs, /about, /legal, /blog, /products, /investors, /press, /leadership) + OpenPageRank, Domain Ages, Web Filtering, IAB Categories, Personas
10 Trading Agents
Python agents using GPT-4o-mini that reason over page-level domain intelligence signals to produce scores, predictions, and recommendations
20 MySQL Tables
Structured output: scores, predictions, anomalies, rankings, and recommendations
5 Dashboards
Enterprise PHP dashboards with charts, DataTables, exports, and real-time monitoring
10 Data Dictionaries
Field documentation with sourcing instructions for real production data

The 10 Programmatic Trading Agents

1. Supply Path Optimization
Analyzes SSP ecosystem health from /api endpoints, /partners networks, /docs quality, /careers hiring trends, and /press recency signals
2. Inventory Quality Scoring
Scores publisher inventory using /about, /legal page detection, OpenPageRank, domain age, IAB categories, and web filtering to flag MFA domains
3. Contextual Targeting
Enriches publisher context from /blog post frequency, /products categories, IAB taxonomy, and persona signals for cookieless targeting
4. Bid Shading Intelligence
Optimizes bid shading by PageRank band — high authority (>7) shades 15-20%, mid (4-7) shades 20-30%, low (<4) shades 30-40%
5. PG Deal Sourcing
Qualifies publishers for PG deals using /about ownership, /investors stability signals, /press media presence, and web filtering safety
6. CTV Inventory Intelligence
Maps CTV platforms using /products ad tier analysis, /partners content deal counts, and /press strategy signals across AVOD/FAST/vMVPD
7. Frequency Capping
Builds cross-domain ownership graphs from /about parent company, /legal entity names, and /leadership page overlap detection
8. Viewability Prediction
Predicts ad viewability by correlating OpenPageRank, domain age, IAB category, ad position, page type, and device type signals
9. Auction Intelligence
Detects domain spoofing (PageRank mismatches), category fraud (IAB vs web filtering conflicts), and suspicious bid patterns
10. Cross-Exchange Arbitrage
Identifies cross-SSP price gaps for the same publisher using /partners and /api signals to verify inventory authenticity

How to Run the Agents

The Python agent script (adtech2_agents.py) accepts command-line parameters to control which agents run and how.

Two integration paths: Use the CSV-based workflow below for batch analysis, or deploy MCP services to connect agents directly to your ad tech platforms for live data pull. Both approaches use the same data dictionaries as the schema contract.

# Populate all 20 tables with realistic demo data $ python3 adtech2_agents.py --populate # Run a specific agent with OpenAI $ python3 adtech2_agents.py --agent bid_shading --run # Run all 10 agents with OpenAI $ python3 adtech2_agents.py --agent all --run # Clear and repopulate $ python3 adtech2_agents.py --clear --populate # Only create tables $ python3 adtech2_agents.py --create-tables
Programmatic Trading AI Agent Platform — 10 Agents • 20 Tables • 5 Dashboards
© 2026 • Powered by OpenAI GPT-4o-mini • MySQL • PHP • Python