Skip to content

Repository files navigation

Brainstorm Security Stack

The Living Case Study

We're building Wiz + CrowdStrike + SentinelOne from scratch.
With AI agents. Live. Come watch and help us test.

Live Dashboard  ·   Brainstorm CLI  ·   BrainstormRouter


What Is This?

A fully public, AI-agent-driven build of a complete MSP Security Stack:

Component Equivalent What It Does
CSPM Scanner Wiz Cloud security posture management, misconfiguration detection
EDR Agent CrowdStrike Falcon Endpoint detection & response, threat hunting
SIEM Engine Splunk/Elastic SIEM Security event correlation, anomaly detection
SOAR Playbooks Palo Alto XSOAR Automated response, incident management

10 named AI agents with persistent identities, domain expertise, and real budgets build this project in the open. Every commit, every PRD, every architecture decision, every code review is visible. Every LLM call routes through BrainstormRouter with full cost and routing transparency.

The Agents

Agent Role Personality
Alex Crypto Engineer Direct, precise, cites RFCs. "HPKE yesterday, not 2030."
Casey API Security Lead Practical, proves vulnerabilities with curl commands.
Sam Compliance Officer Methodical. "If it's not in the evidence ledger, it didn't happen."
Morgan DevOps Engineer Pragmatic. "If it doesn't pass CI, it doesn't exist."
River Risk Analyst Analytical, quantifies everything. Risk matrices, STRIDE scores.
Jordan Auth Architect Systematic, draws sequence diagrams. RFC numbers, not hand-waving.
Taylor QA Engineer Proves bugs with failing tests. "The test is the argument."
Sage Product Manager Clear PRDs, explicit non-goals. "Shipping is a feature."
Quinn Architect Deliberate, ASCII diagrams. "What happens when this fails?"
Avery Frontend Engineer Visual, user-focused. "Every pixel should earn its place."

Each agent has a SOUL.md defining their identity, worldview, voice, and contradictions. They speak like real team members, not chatbots.

Why?

The security stack is impressive. But the real product being demonstrated is how it's built.

  • "This PRD was generated by Sage, routed through BrainstormRouter to GPT-5.4 ($0.06)"
  • "This code review was a 3-agent consensus: Opus disagreed with Gemini, Sonnet broke the tie ($0.14)"
  • "Total cost to build CSPM scanner: $4.23 across 847 routing decisions"

Every metric, every agent check-in, every artifact reinforces: BrainstormRouter makes AI development measurably better, cheaper, and more reliable.

The Open Research Angle

This isn't a proprietary product. It's an open research project:

  • Apache 2.0 — fork it, learn from it, contribute to it
  • Every routing decision logged — "which model is best for security code review?"
  • Every cost tracked — "what does it actually cost to build enterprise software with AI?"
  • Every failure documented — "where do AI agents hallucinate in security-critical code?"

Follow Along

Built With

License

Apache 2.0

About

The Living Case Study — AI agents building Wiz + CrowdStrike + SentinelOne from scratch, live, 24/7. Powered by Brainstorm CLI + BrainstormRouter.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages