<aside> πŸ’‘ Use case: these are the tools that do the labour. Claude thinks and writes, LangGraph runs the agents step by step, PostgreSQL with pgvector remembers, and Midjourney, ElevenLabs and Whisper handle images, voice and audio.

</aside>

<aside> πŸ”— How it connects: a first build runs on Claude Code alone. Claude is the brain, the agent files in .claude/agents/ are the agents, and Markdown files in brand-os/, drafts/ and logs/ are the memory. LangGraph and pgvector come in when the team grows: when you want agents running on a schedule without you opening Claude Code, or when there is more history than a folder of Markdown files can hold.

</aside>

Claude

<aside> πŸ€– Its job here: the reasoning. Claude reads the signals, scores what matters, writes the briefs and drafts every post, caption, article and review reply. In Claude Code, every agent file runs on Claude.

Used by: the orchestrator and every department, most directly Content engine (drafts) and Reviews & reputation (reply drafts), plus every playbook.

From its homepage: "The AI for problem solvers"

</aside>

Claude homepage

Claude homepage

LangGraph

<aside> πŸ•ΈοΈ Its job here: the agent runner. It runs multi-step agents as a set of steps with clear hand-offs, so the six moves (collect, score, brief, draft, schedule, approve) can run on their own outside a Claude Code session.

Used by: the orchestrator, once you move the system off your laptop.

From its homepage: "Balance agent control with agency"

</aside>

LangGraph homepage

LangGraph homepage

PostgreSQL

<aside> 🐘 Its job here: the memory store. A database that keeps every topic covered, every draft, every result and every customer, so the agents can look back before they act.

Used by: the orchestrator and every department, once Markdown files are no longer enough.

From its homepage: "PostgreSQL: The World's Most Advanced Open Source Relational Database"

</aside>

PostgreSQL homepage

PostgreSQL homepage

pgvector

<aside> 🧲 Its job here: the PostgreSQL extension that stores embeddings. An embedding turns a piece of text into numbers that capture its meaning, so an agent can ask "have we already said something like this?" and find it even when the words are different.

Used by: Content engine and the orchestrator, to avoid repeating topics and to find past posts that worked.

From its homepage: "Open-source vector similarity search for Postgres"

</aside>

pgvector homepage

pgvector homepage

Midjourney

<aside> πŸ–ΌοΈ Its job here: visuals. It makes the images for posts and ads from a written prompt.

Used by: Creative studio.

From its homepage: "An independent research lab exploring new mediums of thought and expanding the imaginative powers of the human species."

</aside>

Midjourney homepage

Midjourney homepage

ElevenLabs

<aside> πŸŽ™οΈ Its job here: voice-over. It turns an approved script into spoken audio for videos.

Used by: Creative studio.

From its homepage: "Bringing technology to life"

</aside>

ElevenLabs homepage

ElevenLabs homepage