References

The reference system is the connective tissue of your project. Whenever you mention a claim, figure, reference-sign element, application section, or prior-art patent in your text, or in chat with the assistant, Nodes-IP turns it into a linked chip that points at the underlying entity rather than a literal number or string. References stay correct as your project evolves: claims renumber, figures reorder, elements rename, and every chip follows.

What can be referenced

The same chip primitive handles five kinds of reference:

Claims: claims in your own claim set (claim 1, claims 2 and 3, any of claims 4–7)
Figures: figures in your project (FIG. 2, figure 3a)
Reference-sign elements: the named parts on your figures (the housing 12, the controller 30)
Application sections: the sections of your draft (inserted from the @-mention picker)
Prior-art patents: patents you've bookmarked into the project (US 10,123,456, EP 1234567 B1)

Every chip renders the same way: a coloured, underlined label, with the kind conveyed by the text it shows (Claim 1, FIG. 2, the patent number). Hover any chip to preview the target. Click it to navigate to the claim, the figures gallery, the section, or the prior-art patent panel.

How chips appear

There are two ways a chip ends up in your text:

Automatic detection. As you type prose, Nodes-IP scans for natural-language patterns and replaces them with chips behind the scenes. In "As shown in figure 2, the housing 12 receives the controller 30", every bold phrase becomes a chip the moment you finish the word. You don't type any special syntax.

@-mention picker. In the assistant chat composer, the claim editor, and the section editor, pressing @ opens a picker that lets you pick a reference explicitly. Type a type keyword to narrow it (@pat, @fig 2, @claim 3). Useful when the exact phrasing in your message doesn't match a natural pattern, or when you want to attach a patent to a question without writing it out longhand. The picker is the only way to insert a prior-art patent chip, since patents have no natural-language matcher.

Smart phrasing for claims

Claim references support several phrasing styles, and the system preserves whichever one you wrote:

Single: claim 1
List: claims 2, 3 and 5 / claims 2, 3 or 5
Range: claims 4–7
"Any of" / "any one of": any of claims 4–7, any one of claims 2, 3 or 5

If you typed a range, it stays a range when the underlying claims renumber, even if the new numbers go non-consecutive. If you typed a list, it stays a list. Legal phrasing is yours to control; Nodes-IP only updates the numbers inside it.

Surviving renumbering

Chips bind to internal IDs, not to the displayed text. That means:

Reorder claims and every claim chip everywhere in the project updates to the new number: in other claims, in the description, in the assistant chat history.
Rename a figure or element and every chip pointing at it shows the new name.
Delete a referenced claim and the chip becomes orphaned, shown in amber with a strikethrough and surfaced in the claim editor's diagnostics, so you know to fix it. Nothing is silently rewritten.

This is the system referred to elsewhere as "smart references that survive renumbering", and it works the same way across claims, sections, figures, and chat. For details on how claim numbering and renumbering works, see the Claims tab.

Editing chips

Backspace into a chip to convert it back to plain editable text (minus one character, the standard backspace behaviour).
Click a chip in chat to open its target. Chips in the assistant's replies are fully clickable, just like ones you wrote yourself.
Copy a chip and paste it elsewhere and the chip survives the round-trip if you stay inside the app, degrading to a sensible label (e.g. "Claim 1", "FIG. 2") when pasted into an external tool.

Where references show up

The system is consistent across every surface that handles text:

Surface	Auto-detect	`@`-mention picker	Notes
Claim editor	Yes	Yes	Multi-claim phrasing supported
Application sections	Yes	Yes	Multi-claim phrasing supported
Assistant chat input	No	Yes	Mentions are first-class context
Assistant chat reply	n/a	n/a	Chips are clickable for navigation

The Preview panel renders the assembled application as plain text, with all reference tokens resolved to their current labels.