bmad-marketing-growth: The Declarative Marketing Department

Moving beyond the mega-prompt to a 14-agent hierarchical orchestration framework for SaaS growth.

A vintage organizational chart where boxes are labeled with code keys, and mechanical hands move files into separate sidecar filing cabinets. — Instead of a single chatbot, bmad-marketing-growth encodes a 14-agent organizational hierarchy into its declarative file structure.

Key Takeaways

Sidecar Memory folders isolate platform-specific data to prevent context drift across the agent network.
A three-tier hierarchy delegates high-level strategy from a central orchestrator down to granular platform specialists.
Deterministic YAML workflows transform open-ended marketing tasks into state machines with mandatory human approval gates.

Beyond the Infinite Context Window

The most compelling aspect of `MatthiasMRC/bmad-marketing-growth` is not that it generates marketing copy. It is how it prevents context drift. Most AI marketing tools rely on a single, massive system prompt that eventually collapses under the weight of its own generated context. This repository solves that through a "Sidecar Memory" architecture.

Every agent in the system maintains a dedicated `_memory` folder. This local storage isolates platform-specific heuristics and previous campaign data. A "Reddit Specialist" agent can remember the exact subreddits it researched last week without cluttering the "Marketing Orchestrator" agent's high-level strategic context.

A high-tech motorcycle with a vintage wood-paneled sidecar overflowing with scrolls and maps. — The Sidecar Memory pattern isolates technical debt and platform-specific knowledge from the orchestrator.

The Three-Tier Command Chain

The repository uses a strict Manager-Worker pattern across 14 specialized agents. This is defined in the configuration files as a hierarchical delegation chain. Tier 1 is the Orchestrator, named Max Growth. Max acts entirely as a router and high-level strategist.

When Max decides a campaign needs social media distribution, it does not write the tweets itself. It passes a highly structured Delegation Brief to Tier 2, the Department Leads. Those leads then fan the work out to Tier 3, the Platform Specialists, who execute the granular tasks based on their specific markdown-defined personas.

The hierarchical delegation pattern ensures high-level strategy is preserved while execution is handled by isolated specialists.

Marketing as a State Machine

The true product of this repository lives in the `workflows/` directory. Rather than relying on open-ended chat sessions, the BMAD module treats a marketing strategy like a deterministic state machine. Tasks are triggered by short codes, such as `LS` for a Launch Sequence.

This 21-day timeline coordinates multiple agents from an initial audit through strategy generation, launch day execution, and post-launch review. Crucially, these YAML workflows include defined Gatekeepers. The AI processes halt and require human approval before moving from strategy to execution, ensuring safety and alignment.

Feature	Traditional AI Marketing	BMad Orchestration
Context Management	Single massive prompt (prone to drift)	Fragmented Sidecar memory per agent
Execution Style	One-shot text generation	Multi-phase YAML workflows with human gates
Scalability	Hard-coded chatbot personas	Declarative Tier 3 specialists added via config