A quick note: this is a living draft — I'm rewriting it as we go, so expect a few rough edges. What you'll find across the site has been largely written by AI in a collaborative process, and it reads as such.
Hey — Sam here. I'm raising a $2–2.5M pre-seed for Calyx: a local-first, agent-native workspace built as a fork of VS Code. Think Notion + Obsidian + Cursor had a mutant baby. This page is the long version of why I think it matters — with links scattered throughout so you can wander into the product wherever you get curious.
What it is
Calyx is a true collaboration environment for people and their agents — a company OS where all your work is legible to AI by default.
It has a familiar, Notion-like interface. But under the hood it's an IDE, because it is one: a fork of VS Code. While the interface looks like your standard database-backed app, everything you touch in Calyx is just a view onto the file system underneath. Tasks, projects, kanban boards, infinite canvas, dashboards — behind the scenes it's all plain text files. And coding agents love working with plain text files.
Every doc, task, meeting, and the data from your connected tools becomes a living, queryable model of your company — alive because your agents are working in it continuously, keeping it current. Not a static snapshot scattered across a dozen pre-AI SaaS tools that can't talk to each other.
See it A walk through the product, screen by screen →Why I built it
I started building Calyx in December 2025 to solve my own problem. I was spending all day in IDEs, vibe-coding with agents — and being non-technical, I got sick of the interface. So I built a simple place to create and review the plans my agents generated, and to track features and bugs on a kanban board.
Then it clicked: oh hot damn, I can build the app I've always dreamed of. And here we are. I didn't set out to start a company. I set out to build a better place to work with agents — and realised it was something a lot of people are going to need.
The real problem: collaborating with agents sucks
Coding agents like Claude Code and Codex are insanely capable. But the tooling around them is primitive and developer-centric, and the whole industry is fixated on outsourcing work to agents rather than collaborating with them in real time.
That's the wrong bet for the people who matter most. Top performers don't want to hand their work to a black box. They want to augment themselves — stay in control, shape the output, keep their hands on the wheel. It's about augmentation, not automation. And there's no good home for that today. The most dedicated people are hacking it together inside code editors or bending Obsidian to their will, and anyone who's tried it knows the experience is rough.
I saw this up close. Earlier this year at Dragonfly Thinking, I worked with our Chair, Sue Brake — ex-CIO of Australia's Future Fund — to build multi-agent investment workflows. I'm a capable builder, but I completely lacked her decades of domain expertise. She was the one who needed to build them. But the only tools available were code editors. Sue made it work because she's relentless. Most domain experts aren't. The lesson stuck: for serious people to actually wield these tools, the interface has to change.
Bring your own everything
In Calyx you work with any agent or model you want. Plug in Claude Code, Codex, Gemini CLI, Droid — whatever — and use your own subscriptions and keys. No lock-in to any provider.
This is a deliberate, load-bearing bet. Every day brings a new model, a new harness, a new state-of-the-art. The best model today won't be the best model in six months, and nobody should have to switch workspaces to switch providers. Plenty of competitors lock you into their agent and meter your usage by the credit — we think that's exactly backwards. Companies already pay for unlimited Claude Code and Codex seats; they don't want to pay twice for inference they've already bought.
What we provide is the calyx — the vessel where it all comes together.
How we're different Calyx vs. Obvious, Patina, Mossnotes & the rest →The structural foundation: it's an actual IDE
Here's the thing competitors can't easily copy. Because Calyx is a fork of VS Code, every user is sitting on top of a full execution environment — not a closed SaaS database with a chatbot bolted on.
That lets people do things a standard workspace app structurally can't: edit video, generate PDFs, spin up a custom dashboard, SSH into a virtual machine to configure their own cloud agents, run a real terminal, wire up live infrastructure — all from inside their workspace. Local-first, Git, and Obsidian compatibility come for free on top of that. It's the same property that makes Calyx viable for enterprise (security, control, your data on your machine) and irresistible to anyone who wants to run a file-based company OS.
The interface looks like Notion. The engine is an IDE. That gap is what makes everything else possible.
The SaaS apocalypse — and mini-apps
A reckoning is coming for SaaS. Apps that charge $20/month for narrow functionality are about to get vibe-coded out of existence. People are already building their own replacements. Need a simple CRM? Vibe it. A social scheduler? Vibe it. An invoice tracker? Vibe it. The bar for "worth paying for" just jumped from "saves me time" to "genuinely impossible for me to build myself" — and that bar rises every month.
Calyx isn't a victim of this apocalypse — it's one of the horsemen. Inside the workspace you can build your own mini-apps, download pre-built ones from a marketplace, and customise anything you get. The crucial difference: these apps live in the context of your work — your projects, your documents, your data — which gives them a network effect a standalone tool can never have.
An illustrative example: a friend runs a construction company. I built him a mini-app for agentic document extraction — his team pulls and tags information from architectural door schedules (the spec sheets listing every door in a building), and that feeds a pipeline that generates quotes automatically. A workflow that took hours now takes minutes. No SaaS product does this. No off-the-shelf tool ever will — it's too niche. But with the right platform, you build exactly what you need.
This is a moat and a go-to-market in one: we're not competing with individual SaaS tools. We're making them optional.
Mini-apps & plugins Email, video, presentations, image — running inside Calyx →More than a platform — an ecosystem
Calyx isn't just one app. We're building companion apps that share the same context. The first is live today: CalyxVoice, a meeting recorder and dictation app — think Granola meets Wispr Flow — where transcripts land directly in your workspace, attributed by speaker and connected to the right projects. Then agents post-process them into tasks, feature plans, emails — whatever you want — automatically.
The magic is shared context. A meeting recorder that already knows your projects, tasks, and notes can file itself in the right place and act on what it heard. Each app in the ecosystem is more powerful because of that shared context — and the whole becomes far more useful than any of these tools in isolation.
Companion app CalyxVoice — your meetings and dictation, in your vault →Shipped and working today
This is well beyond an MVP. Calyx has been my daily driver for months. What's live right now:
- Rich Notion-like editor over plain markdown
- Infinite canvas with a built-in AI assistant
- Dataviews — files as tables, kanban, or cards
- Project & task management
- CalyxVoice meeting recorder + dictation
- Agent-first terminal — any CLI, or the built-in agent
- Agent management UI — skills, subagents, cloud agents
- A platform CLI your agents can drive
- AI-generated reports — briefs, summaries, news
- BYO agents & keys — no provider lock-in
Current state is a local-first beta; I'm building cloud sync and credential management now. Technical users can already connect their own services via MCPs, APIs, and CLIs.
Traction
No paying users yet — by design. I'd rather launch with real depth than ship a thin MVP and immediately be in a race with copycats. The goal is to come out of the gate with enough of a head start that catching up takes real time.
The strongest early signal is Dragonfly Thinking — the strategic-intelligence firm where I'm AI lead, with government and corporate clients. We're exploring three uses of Calyx, and the demand for the third is real and recurring: Dragonfly's clients keep asking for an interface to work with its agentic strategic-intelligence system, and Dragonfly has no appetite to build and maintain a platform itself. These aren't small names — the UAE government, pilots with Commonwealth Bank, discussions with Australian super funds.
The early signal Production work, white-label demand, and what I see in the courses →Why "living" is the word that matters
Calyx is a living company operating system — and "living" is doing real work here, not decoration. Agents, and the companies they run on, are cybernetic systems: they persist by sensing their environment, modelling it, and acting to keep that model true. We have mature sciences for exactly this — the Free Energy Principle, cybernetics, complex-systems theory — and their principles aren't a metaphor we apply afterwards. They translate directly into how Calyx is built.
That shapes the architecture on both sides. It changes how we build the agents — what belongs in their context, how they weight attention, when to act versus observe. And it changes how the company OS itself behaves: a living system that mirrors your organisation and keeps itself true by sensing (your connected tools), updating (the vault), and acting to close the gap between the two. Done right this doesn't just make the agents better — it makes the company better: a tighter loop between what the organisation believes and what's actually true.
It's also why the architecture is what it is. Like any intelligent system, agents need low-entropy, legible environments to work in — to read their surroundings accurately and act predictably. Plain files, explicit structure, deterministic folders — not aesthetic choices, but the conditions under which agentic reasoning actually works. And the more of your company that lives inside this system, the more capable your agents become.
This is the real moat, and it's a technical one. What makes Calyx so powerful and useful for a team isn't any single feature — it's that the whole thing behaves like a living system, built on top of a serious body of science (the Free Energy Principle, cybernetics, complex-systems theory) that almost nobody else in this space is even thinking about. Competitors can copy an interface in a weekend. They can't copy a continuously-reconciled, living picture of your company — and they can't shortcut the years of thinking about adaptive systems that shape how we build it.
Agents are a commodity. The living system is the asset — and it's the part that compounds.
Team
Right now, it's just me — a solo founder, actively looking for a technical cofounder with deep infrastructure, agentic-sandboxing, and security expertise. And I think I'm the right person to have started this, because it's a near-perfect alignment of my skills and interests. I know the personal-knowledge-management space about as well as anyone — Roam in the early days, then Logseq, Obsidian, Notion, Coda. At every company I've worked at I've been "the Notion guy" who built the operating system everyone ran on, and I've done the same on a consulting basis paired with AI workflows.
Being non-technical in the traditional sense is an advantage here: I know what an ordinary knowledge worker actually wants from an AI workspace, because I am one. Building with the frontier tools every day — and teaching others to use them professionally — has shown me exactly where people get stuck.
There's a second half to it. I studied philosophy with honours, focused on the Free Energy Principle, thermodynamics, and complex systems — and it isn't background colour. Agents and companies are cybernetic systems, and these disciplines give concrete principles for how to build both: how we architect the agents, and how we shape the user's evolving living context. It directly informs the product. I'm looking to build with people who think the same way — real depth in the Free Energy Principle, active inference, and cybernetics.
Track record What I've built before — and what people say →How we make money
Five revenue streams, layered as the ecosystem grows:
The workspace layer for knowledge work is one of the largest software markets that exists. The upside is enormous.
Where this goes
The raise
I'm raising a $2–2.5M pre-seed to put real fuel behind this. The market is moving fast and the cost of copying drops by the day, so product-led growth alone won't cut it. The money goes to three places: salaries for the rare generalists who can build, ship, and explain; AI tooling and infrastructure; and marketing — a portfolio of AI-generated content plus creator partnerships, educational content where Calyx plays a starring role.
As you can see from the product, I can build. The right technical cofounder takes it where it needs to go.
Want to go deeper?
Explore the product, see how we're different, or just get in touch.