Topic guide: Company brain
Shoreline Topics Company brain
TOPIC 01 Recurring theme · All episodes

The company brain, mapped.

Every Shoreline episode comes back to the same idea: AI gets useful when it can read your business. This guide collects the definition, the episodes, the best timestamps, the tools, and the open questions about building shared company memory.

4 episodes 7 timestamps 6 tools

The definition

A company brain is the organized memory a business keeps so AI agents can do real work: contacts, calls, decisions, brand voice, pricing rules, project history, and the permissions that govern what an agent may read or change. It is not a document dump. The difference between a folder of files and a company brain is structure — required fields, sources of truth, freshness rules, and a review gate before anything writes back.

Shoreline cares about this topic because it is the spine of the whole show. Episode 1 builds the first version out of Notion, CRM records, and agent lanes. Episode 2 turns it into an operations layer that a sales rep can talk to from plane Wi-Fi. Episode 3 asks whether a big enough company brain ends app switching entirely. Episode 4 reframes it as the context constraint. Episode 5 imagines workspaces organized around projects instead of app boundaries, and Episode 6 adds the missing trust layer: QA harnesses that turn recurring failures into repeatable checks.

Related toolsThe memory stack
Cloud memory

Notion

The show's default home for collaborative company memory: notes, CRM-like databases, transcripts, and agent-readable context.

Open Notion AI
Local memory

Obsidian

The local-first counterpoint in the Notion versus Obsidian discussion — personal knowledge that never leaves your machine.

Open Obsidian
Protocol

Model Context Protocol

The shared language for plugging tools, resources, and company systems into AI apps so the brain is actually reachable.

Open MCP docs
Personal agent

OpenClaw

The persistent assistant the hosts use as the interface to the brain — reachable from Telegram, connected to the real workspace.

Open OpenClaw
Agent framework

Hermes Agent

A self-hosted agent framework reference for memory, skills, messaging surfaces, and scheduled work on your own machines.

Open Hermes Agent
Capture surface

Telegram

The lowest-friction way into the brain: voice memos and quick messages that agents turn into structured records.

Open Telegram Bot API
Open questionsStill being argued
Question 01

One universal agent, or many narrow ones?

Episode 3 frames the fork: as context windows grow, does the company brain feed a single agent that moves across support, engineering, and operations — or does review burden keep agents in narrow lanes?

Question 02

How fresh does memory have to be?

Old strategy docs and stale customer notes quietly poison an agent. The show's freshness rule is easy to state and hard to operate: who owns expiring the brain's bad memories?

Question 03

Where does the soft context live?

Episode 4 points at everything that never gets written down — calls, reading, in-person conversations. Capturing it is the highest-leverage gap, and nobody has a clean answer yet.

Question 04

Who reviews the brain's writes?

From wiped CRM tags in Episode 1 to bulk edits at scale: as more agents write to shared memory, the human approval gate is still the part of the system nobody has automated away.

A company brain is not a document dump. It is the organized memory an agent needs to understand roles, systems, values, customer context, and what it is allowed to change.

Shoreline Ep. 3 · key takeaway