ShadowFounder War Room

multiple agents working for you
your co founder
DNA for your brand
AI co founder with strategy tailored to your brand or niche
Creative concepts to beat the competitors

Inspiration

E-commerce founders are drowning in noise. To win today, you need the strategic depth of a CMO, the data analysis skills of a forensic scientist, and the creative output of a Hollywood studio. Most founders can't afford expensive agencies, and most "AI tools" just spit out generic copy.

We wanted to build something "military-grade" for the underdog. We asked: What if you could replace a 10-person growth team with a squad of autonomous AI agents? We built Founder War Room to be the ultimate decision engine—consolidating intelligence, strategy, and high-fidelity video production into a single interface that operates in seconds, not weeks.

Who is it for?

E-commerce founders operating their own online stores who aim to scale rapidly. The solution is designed for teams with constrained budgets, limited manpower, and tight timelines—ensuring that resource limitations do not become a barrier to growth.

Social Impact

We believe this platform will significantly increase the success rate of new and emerging entrepreneurs who would otherwise struggle due to the lack of large budgets or teams. By lowering the barriers to execution and scale, the tool empowers founders to build sustainable businesses with limited resources. Entrepreneurship will no longer be reserved for the elite or heavily funded companies. This platform democratizes access to growth, creating a level playing field where anyone with ambition and vision can build, scale, and succeed—ultimately contributing to stronger economies and more resilient societies.

What it does?

Founder War Room is a multi-agent orchestrator that acts as an automated Co-Founder. You simply enter your store URL and category, and it deploys 13 specialized agents to:

Decode Brand DNA: Instantly extracts visual identity, brand voice, and top products.
Spy on Competitors: Uses Google Search Grounding to find competitors and mine customer reviews for "Opportunity Gaps."
Formulate Strategy: Uses Gemini 3 Pro's Thinking capabilities to draft a tactical 30-day execution plan (CRO, Ads, Email).
Video Forensics (The Deconstructor): You can upload any viral video, and our agents break it down frame-by-frame to reverse-engineer why it worked (pacing, retention hooks, psychological triggers).
Creative Studio (Veo + Imagen): It doesn't just give advice; it does the work.
- It generates high-CTR marketing images using Gemini 3 Pro Image.
- It writes scientifically-backed scripts based on forensic data.
- It turns those scripts into cinematic video ads using Google Veo, handling the prompt engineering automatically.

How we built it?

We built the frontend with React 19, TypeScript, and Tailwind CSS for a "Dark Mode" tactical aesthetic. The core intelligence is powered by the Google GenAI SDK (@google/genai).

We orchestrated a pipeline of 13 specialized agents:

Gemini 3 Pro Preview: The heavy lifter for logic, coding, and the "Deconstructor" video analysis. We utilized Search Grounding for the Intel Officer agent to ensure data is real, and Thinking Config for the Strategist agent to ensure plans are logical.
Gemini 3 Pro Image: Used by the "Nano Banana" agent to generate consistent product photography.
Veo 3.1 Fast: The engine behind our video generation.

The "Veo Pipeline" (Agents 10-13): We created a unique chain for video:

Agent 10 (Script Engine): Writes the script based on forensic data.
Agent 11 (Product Injection): Safely injects user product details/prices into the script.
Agent 12 (Prompt Builder): Translates the script into a technical "camera-ready" prompt specifically optimized for Veo's architecture.
Veo Studio: Renders the final asset.

Challenges we ran into

Orchestrating 13 Agents: Managing the context window and passing state (like "Brand DNA") between agents without hallucination was difficult. We solved this by defining strict JSON schemas for every agent output.
Controlling Video Generation: Getting consistent results from video models is hard. We realized we couldn't just ask for "a video." We had to build Agent 12, a specific agent dedicated solely to translating business logic into "cinematic" prompts (lighting, camera angles, physics) that Veo understands best.
Product Consistency: AI creates great images but often warps products. We implemented a reference image pipeline where users upload their product, and we pass it to the model to preserve identity.

Accomplishments that we're proud of

The "60-Second" Workflow: We successfully reduced weeks of agency work (competitor research + strategy + creative drafting) into a process that takes less than a minute.
The "Deconstructor": We are particularly proud of the video analysis agent. It doesn't just "see" the video; it understands causality—linking a visual cut to a psychological trigger (like "Pattern Interrupt").
Full Veo Integration: Successfully bridging the gap between a text script and a generated video file in a user-friendly modal.

What we learned?

Specialization > Generalization: One giant prompt fails. A squad of small, hyper-specialized agents (e.g., one agent only checks for safety claims, another only writes visual prompts) produces significantly higher quality output.
Thinking Budgets Matter: For the Strategy Agent, enabling the "Thinking" budget drastically improved the quality of the 30-day plan. It stopped giving generic advice and started "reasoning" through the specific market constraints.

What's next for Founder War Room?

Live Audio War Room: Integrating the Gemini Live API to allow the founder to have a real-time voice strategy session with the "Tactical Advisor" while driving or commuting.
Direct Connection with Ecom platforms: connecting with Shopify and similar stores and analyzing data for better decisions.
Direct Ad Publishing: moving from "Simulated Push" to actual API integrations with Meta and TikTok to publish the Veo-generated ads directly.
Multi-Shot Video Stitching: Currently, we generate single clips. The next step is using ffmpeg.wasm in the browser to stitch multiple Veo generations together into a full 30-second spot with audio.