My Eight Bit

The Project

My Eight Bit started as a client portal and grew into the operational brain of our studio. It's the system Don and I use to run Eight Bit Studios, and the same system our clients log into for real-time visibility into their builds.

It pulls knowledge out of Slack, Gmail, Zoom, our repos, Jira, Stripe, QuickBooks, and HubSpot. It structures all of it into a single knowledge graph and exposes it to the AI agents we use every day.

Why a Company Brain

Tom Blomfield at Y Combinator recently published a Summer 2026 RFS called Company Brain. His argument: AI models got good fast, and the new bottleneck is domain knowledge. Every company runs on know-how scattered across heads, threads, tickets, and databases. AI agents can't operate that way. They need a structured, living map of how the company actually works.

We've been building exactly that for over a year. My Eight Bit is what it looks like in practice.

What's In the Brain

• Knowledge Graph - Accounts, apps, tasks, milestones, action items, decisions, contacts, and the relationships between them
• Meeting Memory - Zoom transcripts processed into recaps with extracted action items and decisions
• Conversation Threads - Slack and Gmail tied to the right account, app, and contact
• Code Activity - Commits, PRs, and issues from our repos linked back to tasks and clients
• Financial Pulse - Stripe and QuickBooks for revenue, invoices, expenses, and cash flow
• Pipeline - Web form intakes, CRM contacts, and HubSpot stage data
• Context Entries - Durable knowledge (decisions, system notes, learnings) created by humans and agents

How Our AI Agents Use It

The brain isn't just a database. It exposes about a hundred scoped actions through MCP, the protocol our AI agents use to call into it.

Nano is our 24/7 AI ops agent. It runs on a Mac Mini in our office, reads from the brain, takes actions, and writes back what it learns. It posts a daily priority report, scores inbound leads, drafts content, audits stale tasks, and opens pull requests when given small code changes.

Claude Code runs in our IDEs and pulls from the same brain through MCP. While we're coding, we can ask what was decided about a feature last week, and it answers from the meeting recap.

Both agents read and write to the same source of truth. The brain stays current because the agents keep it current.

The Night Shift

The part that surprises people: most of the work happens while we sleep. The agents run on their own schedule, reviewing the studio's work, keeping the books and the pipeline current, and flagging what needs our attention. By the time we sit down with coffee, the morning briefing is already waiting.

And because agents fail in quiet ways, the system watches itself. When something breaks overnight, it files its own bug report.

Human in the Loop, By Design

Handing an AI agent your company brain raises an obvious question: what stops it from doing something dumb at 3 a.m.? Our answer is gates. Agents propose, humans approve. New knowledge, drafted messages, and code changes all land in review, never straight in the wild.

The result is a system we actually trust. Boring on purpose. The agents handle the volume, and the judgment calls still run through Don and me.

Client Visibility

Most agencies send weekly status emails that are outdated by the time you read them. My Eight Bit flips that model. Clients log in and see their slice of the brain: their app dashboards, milestones and roadmap, meeting recaps, action items, and signed agreements. Everything is live, scoped, and up to date.

Tech Stack