Skip to main content
Back to projects
LiveTool

Seer

The Operating System for a Solo Founder

A persistent operating system that sits above every project I run: memory, governance, and enforcement for AI-driven development. It started as a knowledge graph that gave AI agents continuity across sessions; it has grown into the eye-in-the-sky that briefs, audits, and holds every project to a shared standard.

The Problem

AI assistants are powerful but stateless, and a solo founder running many projects in parallel cannot be the memory, the auditor, and the quality gate for all of them at once. Context gets repeated, decisions get lost, and nothing enforces the standards a team would hold each other to.

The Solution

A structured graph of entities, decisions, tasks, goals, and learning moments, served to AI agents over MCP with a tiered session briefing. On top of it: a governance layer with an engineering-standards catalog, launch-readiness boards per project, an enforcement harness of hooks and gates synced into every repo, and autonomous health checks that watch the whole portfolio.

The Story

Where it started

After months of working with AI agents across multiple projects, the same problem kept appearing: every new session was a blank slate. Decisions got lost, context had to be re-explained, and agents would repeat mistakes they had already been corrected on. It began as the Founder Knowledge Graph, built to give agents persistent memory. Then a second realisation landed: memory alone does not keep quality up. The system needed to audit, gate, and enforce, not just remember. That is when it became Seer.

Where it is now

Live and in daily use across every project. Agents get a tiered briefing at session start with real-time cross-project awareness. The governance layer scores each project against a central standards catalog, generates committed launch-readiness boards, and syncs an enforcement harness (adversarial-review gates, secret scans, git hygiene, pre-push test walls) into each repo. Autonomous checks run around the clock and report back without being asked.

Where it's going

Deeper dashboards for judgment calls that genuinely need the founder, richer derived views over the graph, and expanding the autonomous supervision so the system surfaces problems before they are questions. Parts of it may eventually be productised for other solo founders.

Key Highlights

Full MCP server for Claude Code and other AI tools

Tiered session briefing system (load only what you need)

Engineering-standards catalog with per-project launch-readiness boards

Enforcement harness: review gates, secret scans, and hygiene hooks synced to every repo

Automated drift detection, codebase auditing, and autonomous health checks

Cross-project dispatcher that recommends what to work on next

Tech Stack

TypeScriptSupabaseMCP ProtocolRailwayAnthropic APIResend

By the Numbers

7 projects audited continuously

MCP-native integration

Governance + launch-readiness boards

Enforcement harness in every repo

Have questions about Seer?

Happy to discuss the product, the architecture behind it, or how I ended up building it. If you are working on something related, get in touch.

This site uses Google Analytics cookies to see which pages get read. Decline and it never loads; the cookieless Vercel page counter runs either way. Details on the privacy page.