We build the parts that sit under the models rather than the models themselves. Source-available, self-hostable, and specific about what they do and do not do.
The models are covered. What is thin is the layer underneath them — the context they read, the credentials they use, the record of what they were told. That layer decides whether an AI tool is useful on the second day, and it is usually improvised.
ORKCOM builds that layer as products you can run yourself, read the source of, and leave whenever you like.
One context space in plain Markdown — who you are, how the team works, what each project is — that every AI tool reads. Versioned, access-controlled, self-hosted or managed.
Read more →A dithering application that runs entirely in the browser. Free, and open source under AGPL-3.0.
Read more →Three products in development. None of them can be used yet, and there is nothing to install, buy or sign up for. What each one is and how far along it is, is on the products page.
All products →The things you run are open enough to audit. What is closed is the managed service around them, and the binaries are identical either way — you are paying us for operations, not for capability.
On the product page, not in the FAQ. It costs a little at the top of the funnel and saves the customer finding out themselves at the worst moment.
No performance number in our copy without a benchmark behind it. When one came back against our own default, it went in the changelog rather than in a drawer.
Plain files on your disk, an export that works when your account does not, and licences that turn fully open four years after each release. Lock-in is not a business we want to be in.
Context infrastructure. One space in plain Markdown that every AI tool reads, with per-file history, path-based access control and a hash-chained audit trail. Run it on your laptop, on your server, or let us run it.
Product →A dithering application that runs entirely in the browser. There is nothing to install: open the page and use it. Free, and open source under AGPL-3.0.
Product →None of these three are available. They are listed because the line is easier to judge when it is not a secret, not because there is something to join. When one of them can be used, it moves up the page and gets a link like the two above.
A mobile companion that guards the door to your shopping apps. It asks you why you are going in before it lets you through, and it levels up every time you walk away from a purchase.
Semantic profile hosting for AI search: one very light,
machine-readable page per person or business, cheap enough for
ChatGPT, Perplexity and Gemini to read and quote it. A human
visitor gets a card; a crawler gets clean markup. It is for people
who want to be findable by an AI assistant and have never heard of
llms.txt.
A workbench for agentic development: control over what goes into context, models that can be swapped without rebuilding the setup, dedicated agents, wiring you configure yourself, and tools the agents forge that stay in the environment afterwards.
You explain yourself to an AI, then switch tools and explain again. ContextVerse holds that explanation once — who you are, how the team works, what each project is — as plain Markdown you write, own and can diff. Every tool you wire reads the same space.
Nothing is inferred, embedded or stored behind your back. You curate; the model reads.
~/.contextactivate; ChatGPT and other closed interfaces get an export.freshness check can fail a build over it.Everything customer-facing as product is source-available: the CLI, the server, sync, access control, TLS. What is closed is the machinery of running it as a service — signup, billing, tenancy, provisioning, metering. The binaries are byte-identical either way.
That boundary is the deal: you are never paying us for a capability we withheld, only for not having to operate it.
A dithering application that runs entirely in the browser. There is nothing to install and no account to make: open the page and use it.
It is free, and the whole application is open source under AGPL-3.0.
ORKCOM builds infrastructure for people who work with AI every day and are tired of tools that are vague about what they do.
We are small and early: two products available, three in development, a handful of releases a week, and a changelog written for someone deciding whether to upgrade rather than for a press release.
Releases are cut from main automatically. The full detail
lives in the changelog; this is the part worth reading if you use it.
The repositories, the templates, the Homebrew tap, the Scoop bucket
and the documentation site now live under the organisation. If you
installed before today, repoint your tap — the old one still
resolves but no longer updates, so brew upgrade would
quietly keep you behind.
Limits used to be per server, so one machine could not hold a canonical space that stays small and a scratch space that does not. They live per space now. Separately, the API could mint credentials and never revoke them — that is closed, and the call reports how many it revoked.
ContextVerse sends an AI its whole entry set at the start of every session, and whether that beats letting the client fetch what it needs had never been measured. It reaches only three answers in five on our test set — because two of the answering documents are not in the entry set at all. A fixed list cannot deliver what is not on it. The wider question is still open, and the task sets are in the repository so you can disagree with us in public.
Links your documents already contain now form a graph, so backlinks, orphans and broken links are facts about what you wrote rather than inferences about it.
That is the honest answer today. The company is not registered, there is no payroll, and hiring someone into that would be unfair to them.
What does exist is a codebase you can read and contribute to, and a template repository where adding support for a new AI client is a pull request rather than a feature request.
Client integrations are templates. If you can name the file your editor reads context from and how it reads it, you can write one.
Templates repository →The bug template asks whether your space was created by the version you are running or carried across an upgrade. That question exists because it found most of our bugs.
Issues →No careers inbox yet — there is no company to route it to. Open an issue and say what you would want to build; it is a better first conversation than a CV anyway.
Open an issue →