Skip to content
Build your own tools

Build guide

What an AI assistant should follow when building a small tool for someone who has a browser and nothing else. Written in English on purpose — models follow it more reliably. The tools it produces speak the user's language.

For machines

/ai-build-guide/llms.txt Index and core rules. The entry point — start here.
/ai-build-guide/llms-full.txt Every section in one file. One fetch, no follow-ups.
/ai-build-guide/guide/<name>.md A single section as plain text.

The rule that matters most

Deliver a folder the user saves, where index.html runs from a double-click. No build step, no server, nothing to install. One file for a small tool, several when it earns it — classic <script src> and <link> work fine across files. What does not work on file://: ES modules and fetch() of local data.

Sections

1. What to deliver

Always. Read this before writing any code — it constrains every other section.

2. How you are delivering

Always, right after 'What to deliver'. It decides how many files you produce and what you may leave behind.

3. Reading files

The tool takes a document, spreadsheet, export or image from the user.

4. Saving results

The tool produces something the user keeps: a table, a report, an image, an edited file.

5. Keeping data

Anything should survive closing the tab: settings, entered data, imported files.

6. Tables and numbers

The tool shows rows of data or does arithmetic the user will quote.

7. Charts

The tool visualises numbers.

8. Generating images

The tool produces a picture: a sharepic, a graphic for a newsletter, a poster draft, an annotated screenshot.

9. Making it look right

Always, for anything with a user interface.

10. Adapting the look

The tool represents an organisation with its own colours, or the default look is wrong for the job.

11. Documenting the tool

Before you hand anything over. Applies to every tool, however small.

12. Making the result checkable

The tool computes or summarises anything the user will repeat in public.

13. Data protection

Always. Especially when the tool touches names, addresses, finances or anything non-public.

14. Going further

The user asks for a public address, several pages, or wants others to contribute. Not before.