AI drafts on the firm's own server

Unpublished inventions are confidential. PatentFlow therefore drafts with a language model on your own server. Disclosures, drafts and claims are not sent to any cloud service. The model writes along the guidelines your firm defines.

01

A model on your server, next to the database

The standard installation includes a local model (Qwen 2.5 14B via Ollama) that runs on the same server as the database or on a separate one in your network. It is not required: the firm can also connect external providers only, or work without AI entirely. No AI runs on a submission before someone at the firm has confirmed it.

Further models can be connected: an OpenAI-compatible server of the firm's own, a provider with data centres in the EU or a cloud provider such as Anthropic. For each client the firm sets which models may process its files. Where a local model is installed, it is the default; an external model is only used when explicitly chosen. Every AI call is logged with model, user and file.

The AI providers page with the local model on the firm's server
02

Drafts along your firm's guidelines

Style, structure and checklists come from guidelines the firm writes and versions itself - per task, office and language, for example an example claim set for the EPO or a glossary for English drafts. They are sent with every request to the model, so the drafts follow the firm's style.

From the disclosure the model writes an outline, the claims, the description with one section per figure, and the abstract, for DPMA, EPO, USPTO or PCT, in German or English. A draft usually takes minutes on a GPU server and considerably longer on a plain CPU server; it runs in the background and the attorney gets a mail when it is done.

The AI guidelines with the firm's drafting rules and example claims
03

Editor and chat

Every draft opens in an editor with autosave and undo. Single sections or claims can be revised with an instruction, and a chat next to the draft answers questions and proposes changes that are applied with one click - dictation included, transcribed in the browser.

Warnings point out missing reference signs, claims whose dependencies do not match, and placeholders still to be completed. The final documents are exported as Word, PDF or Markdown in the office's format.

The editor with the description and the chat next to it
04

Claim tree and drawings

The claim tree shows which claim refers to which; renumbering keeps the references right. The figure tool turns sketches, photos and CAD screenshots into black-and-white drawing sheets in the office format and places the reference signs; the vision model can suggest them.

The claim tree with dependencies

Translations with a terminology list

For a filing in another language the model first proposes a terminology list, which the attorney corrects; the translation then keeps to it. A translation can also be done by copy and paste with a tool of your choice, block by block.

Replies to office actions

For an office action, the model reads the objections from the assigned office mail, the uploaded PDF or pasted text: the kind of objection, the legal basis, the claims concerned and the cited documents. For each objection it proposes an argument, claim amendments and the basis in the application as filed. It works from the claims of the chosen version, the description and the claim charts of the file.

A suggestion that denies a cited document a feature is marked as unverified as long as there is no claim chart for that document. The accepted objections become a Word letter to the office, which is kept in the file. On request, the amended claims become a new version of the draft. Without AI, the objections are entered by hand.

Word add-in

The add-in opens PatentFlow in a task pane in Word and links the document to a file. It signs in with the firm's login. Claims, description, abstract or the whole application of a draft version, and the letters of the file, are inserted at the cursor.

The reference sign check covers the whole document. It shows one sign used for different parts, one part with different signs and differences from the draft's list of reference signs. The model revises a selected passage following an instruction; the suggestion replaces the selection only when someone accepts it. The installation file for the firm's IT is under Settings → Integrations.

Connecting AI assistants (MCP)

The Model Context Protocol (MCP) connects AI assistants such as Claude to PatentFlow. The assistant can search clients and files, read files, drafts and claims, list deadlines, record time entries and create tasks. The interface calls the same routes as the web app, so rights, visibility and the audit log apply unchanged, and every tool call is logged.

Three gates protect the connection: the firm switches it on as a whole, each person signs in with their own access, and each client must be released for it. Sign-in uses OAuth 2.1 with a consent page in PatentFlow; clients without OAuth can use personal tokens. The server must be reachable over HTTPS for this. The provider of the assistant processes the answers; the release per client controls which data can reach it.

A demo with an example from your firm.

We show you PatentFlow with a case of your choice and then plan the installation.