Comparisons

FlowLeap vs IPRally: Retrieval Engine or Deliverable?

FlowLeap Team·
#comparison#iprally#patent-search#prior-art#patent-ai-agent
Editorial illustration of a patent citation network beside a report with attached source references.

Disclosure: FlowLeap is our product. Every statement about IPRally below comes from IPRally's own website or the cited source, fetched on 14 September 2026, and is cited in this page's sources. IPRally's retrieval engine is the benchmark we measure our own search against: its graph neural network is trained on millions of examiner citations, and it is one of the very few vendors in this market that publishes a real evaluation rather than an adjective.

TL;DR

IPRally is a search-first platform that has grown into workflows. In 2026 it added a novelty agent, an Invalidity Agent, Analyst AI over patent sets, and a drafting product called Protection. FlowLeap starts at the other end: from the deliverable an attorney signs, with retrieval as one step inside it. If your bottleneck is finding art in a crowded technical field, look at IPRally first. If your bottleneck is producing the memo, chart, or response with the evidence attached and the matter on your own disk, that is the job FlowLeap is built around.

What each one is

IPRally is an AI-native patent search platform. Its proprietary graph search reads technical structures and functional relationships in both the query and the document, and the graph neural network under it is trained with millions of patent examiner citations. On top of that engine sit the agents: upload a disclosure in any format and Agent returns a structured novelty or invalidity report with an interactive feature and claim chart, reasoning and citations per assessment, and a pre-populated IPRally project, without Boolean queries. Its three stated pillars are Search, Intelligence, and Protection, the last covering innovation support, invention disclosure handling, drafting, and office action work.

FlowLeap is an agent workspace. It is a desktop app for macOS and Windows built on a VS Code fork, whose agent searches USPTO and EPO patent data, including WO publications through EPO OPS, alongside scientific literature; it never invents patent numbers or claim text, and every conclusion links back to the passage behind it. A matter is a folder of files on your machine, and the work record of queries, tool calls, sources and stated limitations ships with the report you hand over.

At a glance

FlowLeapIPRally
ScopeDisclosure, prior art, claims, office action, FTO, invalidity, infringement, landscapeSearch, review, monitoring, novelty and invalidity agents, portfolio intelligence, drafting in preview
Where the matter livesFiles in a folder on your machineIPRally cloud projects
Pricing39 per month or 199 per year, publishedNot published; Individual and Team plans
Trial7 days, no credit card3 days, company email required
Model costYour own provider key, no markup and no meterIncluded in the subscription, agreed per customer
Agent-harness accessFull tool set through flowleap mcp and the CLISearch as an MCP connector, in preview, free with a licence
Data sourcesLive EPO and USPTO calls, WO, literature, legal texts, 170M publications for analyticsApproximately 125M publications via IFI Claims, legal status in ten jurisdictions, no non-patent literature yet
Security and certificationsPublic per-feature data-handling page; no ISO or SOC 2 certificateISO 27001 certified, GDPR compliant, blue-chip customer base

The IPRally rows come from its pricing page, its trial page, its agent pricing statement, its MCP announcement, its help-centre data coverage article, its webinar answer on non-patent literature, and its homepage. The FlowLeap rows come from our pricing page, our data-handling page, the MCP page, and the CLI page.

Where IPRally is strong

Retrieval is the moat, and it is a real one. IPRally evaluated its agent on 50 multi-claim cases with examiner novelty citations as ground truth and reported 14 percent better recall, 11 percent better precision, and roughly 2.7 times more unique documents at the same search depth than a single search. Very few vendors in this category publish a number at all, and on a pure novelty sweep in a dense technical field we would expect IPRally to surface art that a query-driven search can miss.

The 2026 release run is fast. The Invalidity Agent, shipped in August, runs a multi-pass search across more than 130 million patents, scores max-per-claim so that one anticipated independent claim flags the patent, and writes the report in the style of the forum, problem-solution for the EPO and the statutory sections for the USPTO. Analyst AI answers a single question across sets of up to 500 patents with paragraph-level citations, and Smart Tags classifies a corpus from a category written in plain English with no labelled training data. Protection, the drafting product, is in preview and launching this month, surfacing the closest prior art in the page as you write, with every AI edit proposed rather than imposed.

Procurement is the other strength. ISO 27001, GDPR, and a customer list that includes Google, Bosch, Nestle, Unilever, Dolby and a national patent office will clear a security review that a younger vendor will not. If your firm's buying process starts with a vendor questionnaire, that matters more than any feature comparison.

Where FlowLeap differs

The price is on the page. One flat plan at 39 per month or 199 per year, with a 7-day trial that needs no credit card and includes free AI models and shared EPO and USPTO data so the trial runs with no keys of your own. IPRally publishes no price on either plan, and agent access is arranged through a 15-minute walkthrough with its team.

The matter is a folder you own. Matters, drafts, figures and generated reports sit on your machine, FlowLeap keeps no copy, nothing you write trains a model, and exactly three features run on our own model accounts: document OCR, the analyst narration on the analytics pages, and the public Try Agent. That is control and portability rather than a claim to be safer; IPRally's own security posture is strong.

The lifecycle past search. Office action responses against US file-wrapper text, freedom-to-operate memos with live national legal events, and element-by-element charting under the all-elements rule are shipped workflows in the CLI and the workspace. At IPRally, office action work is stated scope inside the Protection pillar, and the FTO agent is tentatively planned for the second half of 2026.

Literature and legal text are first-class. The agent searches scientific literature alongside patent data, and the CLI reaches the MPEP, the EPC and the EPO guidelines as full text rather than summary. IPRally has no non-patent literature support today and puts it on the roadmap for probably 2027.

Live office calls, not a pipeline. FlowLeap queries EPO OPS and USPTO ODP at the moment you ask, and the work record shows the query and what came back. IPRally's corpus comes from IFI Claims and new documents become searchable typically within 21 days of publication, with legal status limited to ten jurisdictions and disclaimed for accuracy.

The whole tool set inside your own agent. flowleap mcp is a local MCP server that puts every backend tool, 32 patent tools at setup, into Claude Code, Cursor, Codex or any MCP client, and the CLI carries 26 commands and 32 skills on the same plan. IPRally's MCP connector exposes Search, in preview and free with an existing licence.

Pricing, side by side

FlowLeap publishes one number and no meter: you pay the plan, you bring your own model key after the trial, and your provider bills you directly with nothing added on top. IPRally publishes none. The Individual Plan carries a Start Free Trial button and the Team Plan a Contact Sales button, with no figure on either.

FlowLeapIPRally
List price39 per month, 199 per yearNot published
How you get a numberRead the pricing pageStart Free Trial or Contact Sales
Trial7 days, no card, no keys needed3 days, company email required
Agent or workflow accessIncluded in the planAgent trial via a 15-minute walkthrough, subscription agreed per customer
Model usageYour key at provider pricesIncluded in the subscription

Who should choose which

Choose IPRally ifChoose FlowLeap if
Your bottleneck is recall on novelty and invalidity sweeps in a dense fieldYour bottleneck is the memo, chart, or response and the evidence behind it
You want a published retrieval benchmark and a graph engine trained on examiner citationsYou want live EPO and USPTO calls with the query and response kept in the work record
Procurement needs ISO 27001 and a blue-chip reference list todayYou want a public flat price and a trial you can start without talking to anyone
Cloud projects and shared collections suit how your team worksMatters must stay as files you own, under your own model key
Search inside Claude or ChatGPT is enough agent accessYou want the full tool set inside Claude Code, Cursor, or Codex

Plenty of practices would sensibly run both: IPRally for retrieval depth, FlowLeap for the deliverable and for the claim work around it. Bring one real matter to each and compare what you are holding at the end of the day.

Related comparisons: FlowLeap vs Patlytics, FlowLeap vs Solve Intelligence, FlowLeap vs DeepIP, FlowLeap vs Rowan Patents.