LeadsLogix AI is built on the WAT (Workflows, Agent, Tools) architecture. 7 specialized agent types coordinate via a DAG scheduler to discover, crawl, extract, enrich, and verify B2B data -- each agent autonomous within its budget, all orchestrated as a single pipeline.
Each agent type specializes in one aspect of the intelligence pipeline. The Orchestrator coordinates execution order and manages dependencies.
Discovers companies via search engines, directories, and registries. Manages rate limiting and query rotation across SearXNG, Bing, Brave, DDG, Mojeek, Yahoo, and Google.
Executes 5-layer scraping hierarchy against company websites. Manages per-company budgets (15 pages, 3 browser renders, 120s) and adapts strategy based on website complexity.
Controls Playwright browser instances for JavaScript-heavy sites. Singleton pool management, anti-detection measures, and budget-capped rendering.
Extracts decision makers using 4-method parsing: JSON-LD structured data, team card detection, heuristic+proximity analysis, and LinkedIn X-ray search.
Discovers social media profiles across 8 platforms with cross-platform linking. Extracts follower counts, activity signals, and bio data.
Runs 8-check email verification, entity resolution, deduplication, and confidence scoring. Ensures data quality before export.
DAG scheduler coordinates all agent types. Reactive execution: SEARCH -> CRAWL+EMAIL -> CONTACT+SOCIAL -> VALIDATION. Manages dependencies and parallelism.
Reactive DAG scheduler manages agent dependencies and parallelism
Stage 1 Stage 2 Stage 3 Stage 4
┌────────┠┌─────────┠┌──────────┠┌────────────â”
│ SEARCH │ ───> │ CRAWL │ ──────> │ CONTACT │ ───> │ VALIDATION │
│ Agent │ │ Agent │ │ Agent │ │ Agent │
└────────┘ └─────────┘ └──────────┘ └────────────┘
┌─────────┠┌──────────â”
│ EMAIL │ ──────> │ SOCIAL │
│ Agent │ │ Agent │
└─────────┘ └──────────┘
Sequential ───────────> Parallel ──────────> Parallel ──────> Final
(discover) (crawl + email) (contact + social) (verify)Markdown SOPs define objectives, tool sequences, schemas, and error handling. Workflows are the source of truth for what the pipeline does and how.
Orchestrator layer reads workflows, validates inputs, selects scraping methods, and triggers enrichment. The agent never scrapes directly -- it delegates to tools.
Python and Node.js execution scripts that perform the actual work: scraping, parsing, extraction, verification. Each tool is stateless and independently testable.
7 autonomous agents coordinate to discover, enrich, and verify B2B data. Set your quality threshold and let the pipeline work until it's met.
Everything you need to know about our platform.
Still have questions?
Our team can walk you through the pipeline, pricing, and your use case.