Guerrilla Marketing MCP Connector for Claude

AI agents asked for guerrilla marketing produce the same 5 things every time: "go viral on TikTok", "create shareable content", "leverage social media", "do something creative", "make noise". This is not guerrilla — it's generic advice with a rebellious label. Real guerrilla marketing has three non-negotiable constraints: zero media cost, precision targeting, and built-in amplification.

The Problem It Solves

AI-generated guerrilla marketing fails on five axes:

• Budget creep — "guerrilla" tactics that quietly require $500 in ads. If it costs money, it's paid marketing.
• Spray-and-pray — "post everywhere" without knowing WHERE the ICP's attention is unguarded.
• Dead-on-arrival — tactics nobody shares. They reach 50 people and die.
• Reckless execution — ignoring that guerrilla lives on the edge of ToS, legality, and community norms.
• Untraceable results — "generate buzz" with no way to prove it generated a single dollar.

These aren't creativity gaps. They're discipline gaps. The agent never asks: is this truly free? Where exactly is the ambush? Why would anyone share this? What could go wrong? How do I prove ROI?

How It Works

Guerrilla Marketing uses 5 Decision Pivots — boolean checkpoints that force the agent to prove a tactic qualifies as genuine zero-CAC guerrilla before recommending it:

1. zeroCostVerified — Is the TOTAL cost literally $0 media spend? Only sweat equity allowed.
2. ambushPointIdentified — Have you found the EXACT place and moment where the ICP's attention is unguarded?
3. viralMechanismDesigned — Is there a built-in reason to SHARE, not just consume?
4. riskAssessed — Have you identified legal, reputational, and platform risks?
5. attributionPathDefined — Can you trace from the guerrilla hit to a revenue event?

The tool validates logical consistency. If the agent says GUERRILLA_APPROVED but zeroCostVerified: false, the tool rejects. If the viral trigger is "go viral" instead of a designed mechanism, it rejects. If the risk profile says "no risk", it rejects.

Why It Works

• Tool calls are obligations, instructions are suggestions. The agent can ignore "make sure it's free" in a prompt. It cannot ignore a schema that demands proving zero cost, naming the ambush point, and designing the sharing trigger.
• The commit pattern. The agent proposes its own verdict, then the server validates it against the pivots. This forced commitment surfaces contradictions the agent would otherwise gloss over.
• Semantic traps. The engine catches 4 categories of anti-patterns: budget language in "zero cost" plans, vague channels as ambush points, wishful thinking as viral mechanisms, and dismissive risk assessments.
• 13 consistency rules. Not just pivot-verdict matching — field quality validation and semantic analysis that catch lazy reasoning at every level.