Strategic Intelligence Agent

Strategic Intelligence Agent is an autonomous multi-agent platform that compresses weeks of competitive analysis into ~90 seconds — and ships real artifacts at the end. It is not an AI assistant that produces text; it's an operating system for competitive strategy that visibly plans, debates, self-corrects, and acts. Eleven specialized agents collaborate continuously. A Sentinel runs in the background, monitoring tracked competitors and spawning missions autonomously when it detects a signal — the user never has to click. A Planner decomposes each mission into a JSON plan tree the UI renders live. A Scout investigates with Gemini grounding (or a cached signal corpus in demo mode), emitting evidence records with credibility scores. Then comes the adversarial debate: Predator, Skeptic, and Opportunist analyze the same signals in parallel from opposing strategic angles. When the Skeptic's doubt exceeds 7/10, the system autonomously revises its plan and re-investigates — the plan tree visibly morphs on screen. A Forecaster produces probability-weighted 90-day scenarios. The Synthesizer weighs every perspective into a verdict with a sales battle card. A Critic then audits the debate and can retract or downgrade weak claims — visible self-correction. Finally an Executor takes action: posts a battle card to Slack, generates an executive PDF briefing, annotates open Salesforce deals, opens a Linear ticket, drafts a war-room calendar invite. These are real artifacts you can open after the demo ends. Built with Google Gemini 2.5 (Pro + Flash), FastAPI with SSE streaming, React 19 + TypeScript + Tailwind, SQLite for persistent mission memory, a custom SVG knowledge-graph visualization, and Docker Compose deployment. Multimodal: drop in competitor PDFs or pricing-page screenshots and Gemini vision feeds every agent.