Each agent is a specialized worker with defined capabilities, resource budgets, and permission boundaries.
Discovers company websites, resolves domains, and identifies target organizations from minimal seed data.
Executes the 5-layer scraping hierarchy across discovered websites with per-company page budgets.
Handles JavaScript-heavy pages via Playwright with singleton pool management and render budgets.
Extracts decision-maker contacts using 4-method parsing: JSON-LD, team cards, heuristic proximity, and LinkedIn.
Discovers and links profiles across 8 social platforms with cross-platform identity resolution.
Runs entity resolution, fuzzy dedup, confidence triangulation, and email verification on extracted data.
Manages the DAG scheduler, routes tasks to specialized agents, and makes recursive enrichment decisions.
The reactive DAG scheduler coordinates agents with dependency-aware execution. Agents run in parallel where dependencies allow.
Each agent type has defined capabilities and resource limits. Browser agents are budget-capped at 3 renders per company. Crawl agents enforce 15-page and 120-second limits.
Every agent action is logged with trace IDs, timestamps, and correlation context. Full decision traceability from input to output.
CAPTCHA detection triggers automatic stop and requires_manual flagging. Critical decisions surface for human review before proceeding.
Operations are classified by blast radius. Destructive actions (data deletion, bulk operations) require explicit confirmation.
Everything you need to know about our platform.
Still have questions?
Our team can walk you through the pipeline, pricing, and your use case.