What is shipping now?

MPEP AI's agentic platform, Cyrus, runs multi-step workflows for prosecution, IPR analysis, prior-art search, and drafting. The May 2026 release added multi-pass prior-art investigation, reusable Skills, and faster patent + Federal Circuit retrieval. Plans, notes, Skills, and the per-matter vault compound into the firm's institutional memory. Every workflow ships citation-grounded — each step is anchored to a primary-source URL or paragraph citation.

What agentic workflows does Cyrus run today?

Cyrus runs prosecution, inter partes review (IPR) analysis, prior-art search, drafting, freedom-to-operate clearances, office-action response, Information Disclosure Statement (IDS) generation, invalidity charts, and opinion letters. Each workflow decomposes into an explicit plan, retrieves from the relevant corpora, and produces work product with every cited source visible. Attorneys can pause, edit, or rerun any step on the record.

What is being built next?

The Up Next list below covers institution-likelihood scoring for Patent Trial and Appeal Board (PTAB) petitions, auto-drafted §103 obviousness responses adapted from each firm's prior wins, foreign-equivalent auto-mapping across EPO/JPO/KIPO families, per-practitioner voice preservation so drafts ship in your partner's actual style, and cross-matter portfolio analytics — realization, allowance, and §103 win-rate views available in one dashboard.

How fast does the roadmap move?

The May 2026 release shipped 10 distinct capabilities across retrieval, drafting, workspace memory, and infrastructure. Each Up Next line below is in active development with the same cadence. New capabilities are scoped to ship as agentic-grounded, citation-verified work product from day one rather than as features bolted onto a legacy product. Book a demo to see what is live this week.

What we ship and what's next.

roadmap.txt

//──────────────────────────────────────────────────────────────────
// MPEP AI · ROADMAP & CHANGELOG
//──────────────────────────────────────────────────────────────────

What's shipping, what's next. Click a line to expand.

#UP NEXT

01.Institution-likelihood scoring
before
Filers can't tell ahead of time which claims are likely to be instituted at the PTAB and which will get denied.
what we're shipping
Each draft petition gets a per-claim probability scored against thousands of historical institution decisions, so weak claims surface before you spend the filing fee.
02.Auto-drafted obviousness responses
before
Practitioners rewrite the same §103 traverse from scratch every time, even when their firm has won the argument before.
what we're shipping
The agent surfaces the strongest prior traverse from your own matter history and adapts it to the current rejection — your firm's voice, half the time.
03.Foreign-equivalent auto-mapping
before
Building the EPO / JPO / KIPO family map for a US application is hours of manual lookup and reconciliation.
what we're shipping
Every foreign family member is auto-linked to the US claim set with one cross-jurisdictional view, including priority chains and translated abstracts.
04.Per-practitioner voice preservation
before
Generated drafts read like a generic AI even when each partner has a distinct citation order, signature line, and traversal style.
what we're shipping
The workspace learns each practitioner's conventions automatically — opinions ship in your voice without prompt-engineering for every matter.
05.Cross-matter portfolio analytics
before
Realization, allowance, and §103 win-rate views require pulling reports from three different systems.
what we're shipping
One dashboard covering every active matter — by attorney, by art unit, by examiner — with the cuts that drive next-quarter pricing.

#SHIPPED · JUNE 2026

01.Cyrus — the agentic workspace
before
Even with fast search and solid drafts, the work stayed scattered — you ran a query in one tool, wrote in another, and copied citations across by hand.
after
Cyrus runs the whole turn in one place: it plans the approach, searches every corpus, drafts in your firm's voice, and grounds each claim with a click-through citation — documents, references, and matter history kept right beside the conversation.
02.Skill creator + installable plugin packs
before
Reusable skills had to be hand-authored one field at a time, and there was no way to bring in vetted skills built outside your own workspace.
after
Describe a workflow and the agent drafts the entire skill for you — then manage them all from one Customize panel, or install curated IP plugin packs in a single click.
03.Every PTAB filing, full-text searchable
before
The IPR module indexed the decisions, but the petitions, expert declarations, and exhibits behind them — the documents practitioners actually mine — weren't searchable.
after
All 1.8M+ documents across 18k+ PTAB cases are now full-text indexed — petitions, institution decisions, declarations, exhibits — every passage retrievable in the same sub-second loop.
04.Per-case citation grounding in CAFC answers
before
In answers spanning multiple Federal Circuit cases, a citation badge could resolve against the wrong case's PDF — right passage, wrong document.
after
Citations are scoped per case end-to-end, on the same pipeline that grounds MPEP answers — verified across live multi-case answers with zero misattributions.
05.Daily PTAB + Federal Circuit ingestion
before
New IPR filings and CAFC opinions landed in the corpus in periodic batches, so the freshest matters could lag behind the docket.
after
Automated daily pipelines pull, classify, and index new PTAB documents and Federal Circuit opinions — the corpus the agent searches is current as of this morning.
06.Brand-signature homepage with live demos
before
The rebuilt homepage still read flat in places — washed gradients, a static closer, demo frames that all looked alike.
after
Gold marker signature through the page, a typewriter hero + rotating capability board, and every demo frame in its own duotone — each mockup verified against the real product surface.

#SHIPPED · MAY 2026

01.New homepage and product showcase
before
The marketing site didn't communicate the agentic-first thesis — visitors saw a generic AI-for-law page and bounced.
after
Hero rebuilt around the actual product surface (chat, plans, citations) with seven product sections that animate as you scroll — every claim grounded in something you can demo.
02.Expert-graded prior-art testing
before
We had no way to measure whether our prior-art picks would hold up against an experienced patent litigator's choices.
after
Side-by-side claim-element evaluation against expert benchmarks runs on every retrieval change, so quality regressions are caught before they ship.
03.Multi-pass prior-art investigation
before
One-shot prior art searches missed combinations a thorough practitioner would find by iterating the query.
after
An investigation arc decomposes the claim, fans out multiple searches, then reconciles candidates into a ranked report — closer to how a senior associate actually works.
04.Reusable plans + workspace skills
before
Every matter started from a blank prompt even when the firm had run the exact same workflow dozens of times.
after
Approved plans get saved as one-click reusable skills — §103 traversals, FTO opinions, IPR petitions, claim-chart builders — each callable from the chat composer.
05.Deep research mode
before
Investigating a single patent meant copying the number into search, then chasing related publications and family members manually.
after
Auto-enrichment from any patent number — family, history, foreign equivalents, cited art — surfaced as part of the research turn, not as a separate lookup.
06.Faster patent text retrieval
before
Full-text searches across the patent corpus took multiple seconds per query, slowing the agent's tool loop.
after
Search backend rebuilt — sub-second hybrid retrieval across the full text and citation graph, with batched hydration so claim charts assemble in a single round trip.
07.Faster Federal Circuit case retrieval
before
Pulling on-point CAFC opinions for a brief took the same multi-second hit as patent search.
after
Opinions moved to a dedicated retrieval index with verified relevance ranking — find the closest-on-point precedent in well under a second.
08.Cross-matter reference store
before
Notes, drafts, and citations lived in separate stores so a reference saved on one matter couldn't be reused on another.
after
Unified reference layer across every workspace artifact — anything you cite once is reusable on every matter you touch.
09.UI polish across the workspace
before
Custom cursor and theme states drifted between modules — light-mode broke in a few places, the topbar carried an unwanted seam, motion was inconsistent.
after
Design system aligned across the agentic workspace — same cursor, same theme tokens, same animation feel everywhere.
10.Hardened production sign-in
before
Dev-only auth shortcuts could theoretically reach a production build, and environment promotion lacked a documented checklist.
after
Production-build gate plus a setup doc that locks the path from development to live deployment.

$join_the_waitlist

Follow updates — email [email protected]

Last updated: June 14, 2026