The category
These are not applications. They are one-off utilities — a page with an input box, some logic, and an output pane — and the effort of hosting them has always been out of proportion to what they are. So they end up as a local file that only one person has, rewritten by the next person who needs it.
What people actually publish here
Publishing one
curl -X POST https://read.botook.ai/sites \
-H "Authorization: Bearer otr_live_YOUR_KEY" \
-F file=@regex-tester.html \
-F expires_in_hours=neverOr let the agent that wrote it publish it directly — it usually just did.
Bookmark it, do not re-derive it
A shareable link means the tool exists for the team, not for the person who happened to generate it. Pin the expiry open and put the URL in the channel topic.
Why not a general host
| A platform like Vercel | Here | |
|---|---|---|
| Unit of deployment | A project with a build | One file |
| Setup | Repository, config, pipeline | One HTTP call |
| Private by default | No | Yes |
| Cost of a throwaway | A project you will never delete | A link that expires by itself |
Nothing wrong with a real platform — it is just heavy for a page whose whole source is 200 lines and whose lifetime is one sprint.
Keep it internal
These tools tend to contain the payloads you were debugging, which are exactly the payloads that should not be on a public URL:
- An unguessable link is the floor: no index, no listing.
- An email-domain gate restricts opens to your company's addresses.
- A password works when the audience is smaller than the company.
The strict policy is a feature here
Single-file tools written quickly love to pull a script from a CDN. The viewer's content policy refuses third-party scripts, which means the tool either works self-contained or you find out immediately — better than discovering a year later that your payloads were being parsed by someone else's bundle.
Updating without moving the bookmark
curl -X PUT https://read.botook.ai/sites/SITE_ID \
-H "Authorization: Bearer otr_live_YOUR_KEY" \
-F file=@regex-tester.htmlSame URL, new version. Nobody has to be told.
Multi-file tools
If the tool grew a stylesheet and three scripts, zip the directory and publish that. The tree
is served with index.html as the entry point.
A catalogue of them
Publish an index page that links to the others and pin its expiry open. Now the team has an internal tools directory that took four minutes to build and no infrastructure at all.
What these tools may not do
Worth knowing before you move a utility here, because it decides whether the thing works at all. The viewer runs under a strict policy, so a tool is limited to what it can do inside the page itself — which is most of them, and not all of them:
| The tool | Works here? |
|---|---|
| Parses, formats, diffs or converts text in the browser | Yes — this is the whole category |
Keeps state in localStorage for the person using it | Yes, per reader |
| Pulls in a library from a public CDN | No. Third-party scripts are blocked; bundle it into the file |
| Calls an internal API from the page | Only if that API allows this origin — and think about whether it should |
| Needs a server, a database or a login | No. There is no server side here to give it |
The rule of thumb is that anything which would work from a local file works here, minus the
file:// quirks and plus a URL you can send to somebody. Bundling a dependency into the
single file is usually a one-line change and makes the tool outlive whichever CDN it was borrowing.
Who else can reach it
Only people holding the link, subject to whichever gate you set. Opens and unique viewers show up in the dashboard, which is usually how you discover a tool became load-bearing for someone else's workflow.
Questions
Can the tool call our internal API?
Only if that API allows the origin. The viewer's policy restricts what the page may connect to.
How long can it live?
Forever — expires_in_hours: "never".
What if it needs a server side?
Then it is an application, and it belongs on an application platform. This hosts the document.