build tools for the way you work with ai.
context pages, handoffs, project trackers, and other small tools that make ai more useful between sessions.
what persistent artifact can help agents and projects continue?
01 / what this isa session forgets. the next one starts from nothing, and you spend the first ten minutes re-explaining a project you were three weeks into. the fix is not a longer prompt — it is an artifact that lives outside the chat, holds what was decided, and can be handed to whichever assistant you open next.
that is what this cluster is for: small durable pages that sit at a url. a context page an agent reads at the start of a session. a handoff that moves a project between two different assistants. a tracker that records what was already decided, so nobody re-decides it on monday. none of them are chat features — they are documents, which is exactly why they outlive the session that made them.
a url is what makes them portable. claude, chatgpt, codex, cursor and grok can all read a page on the open web, and none of them can read the transcripts of the others. the page becomes the shared memory between tools that have no other way to share anything.
publishing is part of the loop rather than a step afterwards. every prompt here ends by telling the assistant to publish the finished page and hand back the url — over an htmlshare connector if you have one configured, and otherwise over the rest api, which needs no account or key. you end the message with a link instead of a file you still have to put somewhere.
questions.
03 / questionscan the assistant publish the page itself?
yes, and the prompts are written for it. with a connector configured the page is permanent and owned by your account. without one, the assistant publishes anonymously over the rest api — no signup, no key — and reports back the url, a private claim link, and the manage token that can edit the page later.
which assistants do these work in?
claude, chatgpt, codex, cursor and grok, as written. the prompts are plain text that ends by asking for one self-contained html file, so there is nothing to install and no plugin to configure — the format is the compatibility layer.
how do i update a context page next session?
republish over the same url rather than making a second page. the manage token from the original publish authorizes that, and an account-owned page is editable from the account — either way the link already sitting in your notes keeps resolving to the current version.
is a context page private?
no. anyone holding the url can read it, and pages are not hidden from search engines. keep credentials, keys and anything under nda out of it — a context page is for decisions, structure and open questions, which is most of what a session actually needs to resume.