Introducing Concepticon
A long story finally has a working ending. The first product in the Thinking Tools Software suite — Concepticon, a desktop app for capturing your team's ubiquitous language as plain-text concept maps — is now available.
Concepticon is a desktop application for capturing your team’s ubiquitous language — the concepts and propositions that describe your problem domain — and managing it the way you’d manage source code: in plain text, in your own files, in version control.
The free Reader is out today. Pro is $50 USD, once. There is no subscription, no cloud account, no telemetry. The proposition file is the source of truth; the concept map is a view of it.
The rest of this post is the why now and the where this came from for the people who care. The download link is at the bottom.
A problem that has always mattered
It has always mattered for a team to be on the same page about what words mean in their domain. Customer. Order. Account. Transaction. These look like obvious words. Six engineers asked to write down what each one means in their corner of the system will return six different definitions — and ship code, day after day, against those subtle disagreements.
The discipline of building the shared model — Eric Evans’s ubiquitous language in the domain-driven design lineage — has been around for two decades. The underlying problem has existed for as long as more than one person has worked on the same piece of software.
In 2026 it matters even more, because your newest team members are AI coding agents — and they need that agreement in an even more accessible form than humans do.
A freshly spun-up Cursor or Claude Code session is, functionally, a colleague who joined yesterday. It has no domain context until something is fed in. The densest plain-text representation of a team’s domain — a concept map of the ubiquitous language — solves the agent’s onboarding problem with the same artefact that solves a human new-hire’s onboarding problem. The format that was a principled choice in 2009 turns out, in 2026, to be the only one that works for the way the work is now being done.
The long road back to this tool
The first version of this tool was a prototype I built in the late 2000s with my friend Simon Harris, while we were inside complex software-architecture problems that were larger than any one person could hold in their head. We came at it from a frustration with diagramming tools: they made you expose all the complexity at once, which made the maps almost impossible to use in practice. The insight from that period — the knowledge captured in a meaningfully complex domain is always unapproachably complex when rendered in a single diagram, so a successful tool must tackle the approachability problem — is the load-bearing reason this work exists at all.
In the mid 2010s, with Sebastian von Conrad, I made the tool publicly available at conceptmaps.io. Same plain-text proposition format the prototype had used. I used the site to promote the ideas on the conference circuit for about three years.
In mid 2024, a Twitter sign-in issue I couldn’t solve on my own took the site down. By that point I was running engineering teams full-time and no longer had the practical technical skill or time to fix it. I let it go.
That Christmas I came across Cursor — and regretted my decision. By 2025, AI coding tools had matured to the point where I could realistically rebuild the application as a single technical operator, without going back to full-time programming. I made one design call up front: rebuild it as a desktop application rather than a web service, so the data could completely belong to the user. That rebuild is what ships today.
What also made the commercialisation possible
Building the product as a single technical operator is one thing; running a small indie software business around it — product decisions, support, marketing, finance, the planning that keeps it moving — is another. That side has been made tractable by a virtual team of AI collaborators, each running on an open-source agent framework built by Norman Noble: agent-framework. The framework treats agents as senior collaborators rather than task runners — persistent memory across sessions, explicit autonomy levels for what each one decides on its own. The fact that this product is shipping commercially today, and not sitting on a hard drive somewhere, owes a lot to that framework.
What this isn’t
Concepticon is not a mind-mapping tool. Mind maps are single-root hierarchies — one central concept, branches radiating out. Concept maps are graphs with labelled edges between any pair of concepts. The structural difference matters: a concept map can capture cross-cutting relationships that a hierarchy cannot represent.
Concepticon is also not a tool for business rules, processes, or architecture diagrams. Concept maps describe what your team’s words mean to each other — they are the foundation underneath all of that other work. If your team needs to capture rules, processes, or arguments as well, those belong in separate tools designed for those shapes. Two of them — Processium for processes and Logicon for the Theory of Constraints Thinking Processes — are next in the suite.
Finally, Concepticon is not a cloud product. Your propositions live on your disk, in plain .md files you can read in any editor. There is no account to create. The Pro version uses AI features that call external APIs at your direction; the Reader uses no APIs at all.
What’s next, and what you get today
I have other products in mind that take the same philosophy — process models, and logic models inspired by the Theory of Constraints Thinking Processes. Together they’ll give your team (human and AI) even more plain-text context for staying on the same page.
For today, the first product is ready:
- The free Reader is for viewing, navigating, and sharing concept maps — no purchase, no account, no email.
- Concepticon Pro ($50 USD, once) is for the people who make them: in-app proposition editing, AI map generation from documents and URLs, and rich concept descriptions.
This first release is for macOS on Apple Silicon. Windows and Linux builds will follow.
If this resonates and you have engineering or product friends who’d recognise the problem, sending them the link is the kindest thing you can do for a small indie shop on launch day. Thank you for reading.
— James