Claims

The Claims page is where you draft, edit, and manage the claim set for your application. Each claim is its own rich-text editor with structural awareness: Nodes-IP understands the preamble, transition phrase, and individual limitations, and uses that understanding to power formatting, references, diagnostics, and assistant edits.

Automatic structuring

You write claims as plain prose. Nodes-IP parses what you type and lays it out in proper claim form automatically:

The preamble (everything up to the transition phrase) is set off as the opening line.
The transition phrase (comprising, consisting of, wherein, etc.) is detected automatically.
The limitations are split at each semicolon and rendered as a lettered list (a), (b), (c) …

You don't need to add the letters yourself, format the indentation, or hand-build a list. Paste in a claim from anywhere (a draft, a granted patent, a generated suggestion) and the editor restructures it on the fly.

List view and tree view

Use the layout toggle in the page header to switch between two ways of looking at the same claim set:

List view shows claims in their numbered order, top to bottom. This is the default and matches how the claims will appear on the printed page.
Tree view groups dependent claims under their parent, so you can see the dependency hierarchy at a glance. A claim that depends on multiple parents appears once under each parent.

Drag-and-drop reordering works in both views. In tree view, dragging a claim moves its whole subtree as a block, so child claims always stay under their parent.

Each claim card is either active or inactive, and only one claim is active at a time. The card you are editing is active — it shows a thin coloured outline and behaves like any text editor. Every other card is inactive and behaves like an object you can pick up:

Drag from anywhere on the card to reorder — no modifier key needed.
Right-click anywhere on the card to open the claim actions menu (add claim below, add dependent claim, focus on chain, copy, show parsed result, delete). Right-clicking an AI suggestion inside the text adds accept / reject at the top of that menu.
Click to activate the card and place the cursor where you clicked. From then on typing, selecting text and the browser's own right-click menu (spellcheck) all work as usual.

Selecting text therefore takes two steps on a card you are not already editing: click it once to activate, then select. Pressing <kbd>Esc</kbd> or clicking away deactivates the card and saves it.

Focus on a claim's chain

Focus on chain, in a claim's actions menu, narrows the list down to one claim's full dependency chain: its ancestors (the limitations it inherits) and its descendants (every claim that would be affected if you edited it). Every other claim is hidden, collapsed into a gap line that shows the hidden claim numbers.

A claim that depends on several parents ("any of claims 1 to 3") pulls in all of them, not just one path. Click a gap line, click show all in the pill at the top of the list, or press <kbd>Esc</kbd> to exit and see every claim again. There's no way to peek at a single hidden claim without exiting: the gap either collapses or the whole set reappears.

Focus on chain is unavailable when the claim set has a broken or circular dependency: with a chain that can't be resolved reliably, Nodes-IP shows every claim rather than guess.

Chain focus is separate from the ?claim= deep links the assistant and write page use to jump to a claim: opening a deep link never hides other claims, only Focus on chain does.

Renumbering

Claim numbers are assigned automatically based on order. When you add, remove, or reorder claims, every claim renumbers in place, including any inline references to those claims elsewhere in your application (see Smart references below).

You never edit a claim number by hand. To change a claim's number, change its position.

Automatic checks

Nodes-IP runs diagnostics on the claim set as you type and surfaces issues in the issues panel and inline on each claim. The checks are structural rules that run on every keystroke. They cost nothing and update without delay.

Each check has a mode under Settings → Claims (reachable from the View menu in the page header), where they're grouped by category in collapsible sections: Always runs it in every jurisdiction, Auto only in matching jurisdictions, Quiet keeps it in the issues panel but hides its marks in the editor, and Never disables it. Diagnostics are advisory and never block you from saving a claim. A few checks are off by default because they cover stylistic preferences or fee notices rather than substantive defects; turn them on in Settings if you want them.

Dependencies

Orphaned references: an inline claim chip points at a claim that no longer exists.
Unlinked mention: a reference sign like (44) or a claim number like claim 1 is written as plain text instead of a linked chip. A chip stays attached to the element or claim it names and renumbers itself automatically; plain text does not, and silently goes stale the next time something is renumbered.
Unknown reference sign: a parenthetical group like (44, 546) mixes a sign that matches a declared element with one that doesn't. Usually a typo or an element you never added to the drawings. A group where nothing matches, such as a year, is left alone.
Broken dependencies: a claim depends on a number that doesn't exist.
Self-dependency: a claim lists itself as a parent.
Circular dependencies: claims form a cycle (A → B → A).
Forward dependencies: a claim depends on a claim that comes later in the set.
Multi-on-multi: a multi-dependent claim depends on another multi-dependent claim (rejected by most patent offices).
Insufficient differentiation: a dependent claim adds no apparent limitation beyond its parent (35 USC 112(d) "must further limit").
Multi-dep claim using "and" (US only): a multi-dependent claim joins its parents with and instead of or; 35 USC 112(e) requires the alternative form.
Category mismatch: a dependent claim's category (system / method / use) differs from its parent's.

Structure

Missing preamble: the claim opens with a transition phrase but has no preamble noun.
No transition phrase: the claim has no recognisable comprising / wherein / consisting of transition.
Mixed transition phrases: the claim mixes transition phrases of different scope, such as comprising (open) and consisting of (closed). Repeating the same transition inside sub-components is fine and is not flagged.
Multiple independent claims of the same category (EP only): more than one independent claim of the same category (apparatus / method / use); EPO Rule 43(2).

Clarity

Relative term: vague qualifiers like substantially, approximately, or about that may be objected to under EPC Article 84 / 35 USC 112(b).
Subjective term: easy, fast, user-friendly, optimal, and similar qualifiers that lack objective measure.
Limiting term: absolute words like must, only, essential, or always, and self-referential phrases like the invention, that courts and examiners can read as narrowing claim scope.
Optional language in claim: preferably, optionally, may be, can be inside a claim creates ambiguous scope; the EPO routinely objects.
and/or usage: A and/or B is ambiguous; use Markush form (at least one of A and B) or split into alternatives.
Ambiguous transition: having, including, or containing as the main transition has uncertain open/closed semantics; prefer comprising or consisting of.
Consisting essentially of: flags consisting essentially of for review; triggers a heightened spec-support requirement.
Means-plus-function (US only): means / step for X-ing invokes 35 USC 112(f) interpretation; the claim is limited to the corresponding structure in the spec. Easy to write without meaning to, so it stays on as a guard.
Undefined acronym: an acronym used in a claim, such as CPU, is never spelled out anywhere in the claims or the description.
Inconsistent acronym: an acronym used in a claim is spelled out more than once with different meanings, such as central processing unit in one place and computer processing unit in another.

Counting

These checks compare the claim set against a limit for words per claim, total claims, and independent claims. Each row in Settings → Claims → Counting has a number field and a single toggle to turn the check on or off.

Leave the field empty and it resolves from your matter's jurisdiction: the USPTO's fee-free thresholds (20 total claims, 3 independent) for a US matter, the EPO's 15-claim threshold for an EP matter (the EPO charges no separate fee per independent claim, so that one falls back to the generic default of 3), and the generic defaults (20 total, 3 independent) for PCT, which has no equivalent claims fee. Type a number to set your own limit; clear the field to revert to the resolved default. The word-count-per-claim limit has no jurisdiction preset and always uses the generic default (250) unless you override it.

Filing more claims than the fee-free threshold means the office charges an excess-claims fee, so the claim count is calculated the way the relevant office counts it: at the USPTO, a multiple dependent claim (one that depends on more than one other claim) counts as that many claims for fee purposes, not one; everywhere else every claim counts as one. When that changes the total, the warning shows both counts.

Word count per claim limit: flags any claim with more words than the limit.
Claim count limit: flags a set with more claims than the resolved limit; names the office and its fee-free threshold when the limit came from a jurisdiction preset.
Independent claim count limit: flags a set with more independent claims than the resolved limit.

The issues panel

The Issues card to the right of the claim list is a grouped summary: one row per check that currently finds something, with a count. The eye button on a check's row hides that check's marks in the editor while its issues stay counted here; click it again to bring the marks back. This is the same as setting the check to Quiet in Settings.

Pending assistant changes are proposals awaiting your review, not issues, so they are not counted here. The issues (and pending changes) themselves live in the editor:

Hover a flagged phrase to read its warning.
Hover a pending change to accept or reject it in place.
Warnings with no specific phrase to point at (structural and dependency checks) tint the claim number instead. Hover the number to read them.

You can collapse the side panels and give the editor the full width. The header shows a count of all open issues: click it to hide the panels, click it again to bring them back. Your choice is remembered. Drag the divider to resize the panels.

Find and replace

Press <kbd>Cmd</kbd>+<kbd>F</kbd> (or <kbd>Ctrl</kbd>+<kbd>F</kbd>) to open the find-and-replace panel in the top-right of the claim list. Type a phrase to see every match highlighted in the editors and a live count across the set. Toggle Aa (match case) or Ab (whole word) to narrow the search, then choose Replace all to apply the substitution to every claim that contains a match. Inline reference chips (claims, figures, elements) are left intact: only the surrounding prose is rewritten. Press <kbd>Esc</kbd> to close.

Statistics panel

A Statistics card above the issues panel shows a live readout of the claim set: total claims, the independent and dependent counts, and the word count across the resolved prose. The claims and independent-claims counts are annotated with their resolved limit, for example "18 / 20 fee-free (USPTO)" or, with an override in place, "18 / 25 (your limit)". The claims count always shows the number of claims in the set; when the office counts differently for fees (a US multiple dependent claim counts as several), the fee-purpose count appears alongside, for example "18 / 20 fee-free (USPTO) · 21 for fees", and the comparison against the limit uses that fee-purpose count, turning the number to the warning color once over it. This is ambient context, not a second warning; the full advisory diagnostic still lives in the issues panel.

Smart references that survive renumbering

When you write claim 1 (or claims 2 and 3, or any of claims 4–7) inside a claim, Nodes-IP replaces the text with a chip that points at the underlying claim, not the number. The chip stays linked even when the claim list is reordered or renumbered, so:

Dragging claim 5 to the top updates every reference to it everywhere: in other claims, in the description, in figure callouts.
Deleting a referenced claim doesn't silently break the link. The reference becomes an orphaned chip (highlighted amber) and shows up in diagnostics so you can fix it.
Phrasing is preserved. If you wrote "claims 4–7", it stays as a range; if you wrote "claims 4, 5 or 7", it stays as a list. You stay in control of legal phrasing. Nodes-IP only updates the numbers.

The same chip system works for figure references (FIG. 2), reference signs from your drawings, and section references (the Background). See the References tab for a full overview of how chips work across every surface.

The assistant can read and write claims

The right-hand assistant panel has full read access to your claim set and can edit it on your behalf. You can ask it to:

Draft new claims: independent or dependent, method or apparatus, in any jurisdiction's preferred style.
Refine existing claims: broaden, narrow, fix antecedent basis, restructure limitations, swap transition phrases.
Audit the set: check for support in the description, surface novel features, suggest dependent claims to add.
Restructure the set: add, edit, and delete claims to group related ones, split an independent claim into separate trees, or rework dependency chains.

Edits the assistant proposes show up as pending changes on the affected claim. You review the diff and accept or reject before anything is saved. The assistant never silently overwrites your work.