About Hark
Hark is named for the Harkness table — a round table with no head, where every seat is equal and every idea gets heard. That's the whole idea behind the product: one home base where your own thinking, your team, and your AI tools all sit at the same table.
The Harkness table
The Harkness table is a teaching method built around a simple piece of furniture: a large oval or round table where a class sits facing each other instead of facing a teacher at the front. No seat is privileged, and the conversation belongs to whoever's at the table.
We kept coming back to that image while building this. Working on an idea today usually means switching between your own notes, a teammate's opinion, and two or three different AI assistants — and each of those tends to act like it's the only one in the room, with no memory of what the others already know. Hark is our attempt to build the table instead: one place where all of that sits together, and where any one of them can add to the record without needing to be the head of it.
Who's building it
We're a small independent studio. Hark started as the tool we wished existed for our own projects — a way to hold a handful of ideas at different stages of readiness, keep the thinking behind each one somewhere durable, and stop re-explaining context to whichever AI tool we happened to be using that day.
We're still using it that way. New features tend to come from a real gap we hit while running our own project pages through Vault, Recon, Blueprint, Fabrication, Live, and Cold Storage, not from a roadmap written in the abstract.
What we're building it for
Hark is the model-agnostic home base for every idea you're building — a hosted MCP hub that lets Claude Code, Codex, Cursor, ChatGPT and Grok share the same project memory, with venture-locked sessions, reviewable, attributed agent memory, repo-native handoff export, a pipeline from spark to shipped, and public project pages.
In practice, that means a project page for each idea, a living Current state that says where it actually stands, a journal that keeps the decisions instead of losing them, and a hosted MCP endpoint so the agents you use — whichever ones they are — can read and write that same record.
What we won't do
We try to be plain about what Hark is and isn't. It doesn't replace a document editor, an issue tracker, or your judgment about the idea itself. It's a narrower, more opinionated tool than most of the software it sits next to, and we'd rather it do that one job well than grow sideways into everything.
You can see how it's put together on the how it works page, browse what people have chosen to make public on Open Studio, or reach us directly at hello@harkstudio.io.
Common questions
- Where does the name Hark come from?
- From the Harkness table — a round seminar table with no head, where every seat is equal and every idea gets heard. It's the origin of the name and the idea behind the product: one table where your own thinking, your team's, and your AI tools' all have a seat.
- Who builds Hark?
- We're a small independent studio. We built Hark because we kept losing track of our own ideas across too many docs, tools, and half-finished chats with different AI assistants, and wanted one home base that didn't play favorites between them.
- Is Hark affiliated with any AI company?
- No. Hark is deliberately model-agnostic — it isn't built or owned by any of the AI vendors it connects to. It talks to Claude Code, Codex, Cursor, ChatGPT, VS Code, Windsurf, Gemini CLI, and headless scripts on equal terms, which is the point of the Harkness table framing.
- How can I contact the team?
- Email hello@harkstudio.io. That reaches the people building Hark directly.
- Where can I see what's shipping?
- The changelog lists what's shipped recently, and Open Studio is the public gallery of project pages people have chosen to share.