Microsoft 365 native · Teams & Word · No IT project required
Legal intelligence
built into the tools
you already use.
Marcella runs inside Microsoft Teams and Microsoft Word. no new software, no new login, no data leaving your tenant. 50-state statutory corpus, treatise citation intelligence, CMMC compliance, and matter memory that compounds with every session
90s
3-hour associate
research question
research question
0
Hallucinated citations —
architecturally impossible
architecturally impossible
50
States. statutory &
treatise doctrine mapped
treatise doctrine mapped
∞
Matter memory —
every session compounds
every session compounds
How It Works
A platform that lives
where attorneys work.
T
Microsoft Teams nativeAsk a research question in the Teams channel you already use
W
Microsoft Word nativeResearch and clause analysis without leaving the document you're drafting
Marcella is an orchestration layer inside your Microsoft 365 tenant. It connects to the statutory corpus, indexes your matter documents, and responds to queries. all without leaving Microsoft. Client data never reaches any external server. Marcella does not train on your queries
Every output is grounded in the indexed corpus. statute, case, or indexed authority cited by name. No hallucinated authority. Every session is linked to the named matter and auditable
What Changes After Marcella
90 seconds vs. 3 hours. Research time cut. You bill the same rate. That gap is pure margin.
You walk into a deposition six months after the initial research. Marcella already knows the matter. No re-briefing
Every contract your client sends is classified GREEN / YELLOW / RED before a senior law pro touches it
Privilege review that used to take a contract attorney three days takes an afternoon. the audit trail is already built
Your CMMC evidence package is assembled automatically at ingest. Nothing to compile before the assessment
Data stays in your Microsoft tenant · Zero training on queries
Four Ways to Access Marcella
Start in the tools
you already have.
T
Microsoft Teams
Ask Marcella in the Teams channel your matter team already uses. Research, triage, and analysis returned in the conversation thread. Matter-keyed automatically.
Get Teams Access
W
Microsoft Word
Clause analysis and research without leaving the document you are drafting. Highlight a clause, ask Marcella, get a citation-backed verdict in the sidebar.
Get Word Access
Slack COMING SOON
Marcella in your firm's Slack workspace. Same grounded research and triage in the channels where your team already coordinates. No M365 requirement for Slack access.
Notify me when available →
Clio
Clio COMING SOON
Marcella connects to your Clio matter records. Triage and research linked directly to your existing matter management system. Matter memory synced with your Clio matters automatically.
Notify me when available →
Pricing
Built for every
practice size.
We are rolling out new capabilities weekly. The price you sign up for today is locked for the life of your contract. Prices increase at full MCM launch.
Marcella Free
$0
Solo · Start today · No credit card
✓ 3 queries per day
✓ Federal corpus — statutes and case law
✓ Document triage — GREEN / YELLOW / RED
✓ 1 active matter · Teams and Word
✗ No AG opinions · No state corpus
✗ No Universal Legal Reference Layer
Free Tier Add-Ons
Build what you need.
Add-ons stack. When the monthly total approaches $399, Marcella Practice is the cleaner choice.
State corpus pack (WAC, RCW, etc.)$40/state/mo
Query pack (50 queries)$15 one-time
Matter memory expansion$25/mo
Bankruptcy corpus$65/mo
Crypto / 1099-DA · AG opinions · and morefrom $30/mo
Marcella Practice
$399/seat/mo
Beta · MAP $599/seat
Per seat · add as needed
✓ Unlimited queries
✓ Choose 6 corpus packs
✓ Bankruptcy · Crypto · Family law · Healthcare · Immigration · 20+ packs
✓ Full matter memory
✓ Secondary case treatment · AG opinions
✗ No Universal Legal Reference Layer
Westlaw avg ~$171/seat · no matter memory
Marcella BOYA
$999/mo
Beta · MAP $1,500/mo
Flat fee · up to 15 users
✓ Flat monthly fee — every user included
✓ All 50 states · full corpus
✓ Universal Legal Reference Layer
✓ All practice areas · specialty corpora
✓ Full matter memory · eDiscovery
✓ CMMC evidence vault · client portal
✓ Runs in your Azure tenant
✓ API access
Saves $198/mo vs 3 Practice seats
Requires M365 + Azure subscription
Marcella Firm
$1,999/mo
Beta · MAP $2,999/mo
16–50 users · contact to configure
✓ Everything in BOYA
✓ 16–50 users
✓ Dedicated configuration session
✓ Priority support
✓ Quarterly review
Marcella Enterprise
Custom
Bar · MCF · MSP · 51+ users
✓ Everything in Firm
✓ Bar association site license
✓ MCF on-premises appliance
✓ Full managed services (MSP)
✓ Custom corpus build · SLA
✓ White-label option
Above Enterprise — The Hardware Tier
MCF Appliance
Marcella Compliance Frame · On-premises · Contact for pricing
For firms requiring on-premises data sovereignty. Matter memory lives locally. Compliance evidence chained in hardware. Client data never leaves the building. Three deployment sizes based on storage capacity and seat requirements.
MCF COMPACT
Mid-market · Up to 50 seats · Entry storage capacity
MCF ENTERPRISE
Large firm · Unlimited seats · High-capacity · Air-gap capable
MCF SOVEREIGN
Government · AmLaw 50 · Custom configuration · Petabyte-scale
Marcella Free activates immediately. Marcella Practice, BOYA, Firm, and Enterprise provisioned within 24 hours of agreement.
Beta prices are locked for the life of your contract. Prices increase at full MCM platform launch.
Beta prices are locked for the life of your contract. Prices increase at full MCM platform launch.
The Case Every Attorney Needs to Know
United States v. Heppner
Your client's AI sessions are discoverable.
In February 2026, Judge Jed Rakoff of the Southern District of New York issued a ruling of first impression nationwide: documents a criminal defendant created using the free public version of Claude. analyzing his own legal defense strategy, after receiving a grand jury subpoena. were not protected by attorney-client privilege or the work product doctrine
The defendant, Bradley Heppner, had used Claude to prepare defense strategies and later shared the AI outputs with his attorneys. The FBI seized the documents in a search warrant. Judge Rakoff ruled them fully discoverable. because the public version of Claude has no confidentiality expectation, trains on user data, and can share inputs with third parties including government authorities. Privilege never attached
Judge Rakoff explicitly noted the outcome might have been different if counsel had directed the client to use an enterprise AI tool. one that kept data confidential and operated within the attorney-client relationship. That is exactly what Marcella is. The ruling is not a warning against AI. it is a roadmap for how AI use can be privileged
How Marcella Makes Both Heppner Problems Architecturally Impossible
HEPPNER PROBLEM 1
Public AI has no confidentiality. sessions are discoverable
Judge Rakoff found Heppner's Claude sessions were not confidential because the free public platform collects user inputs, trains on them, and can share them with third parties including the government. No privilege attached
→ Marcella runs inside the firm's M365 tenant. It does not collect your queries. It does not train on your data. It cannot share your data with anyone outside your tenant. The confidentiality element that Heppner lost is preserved architecturally
HEPPNER PROBLEM 2
Client used AI without attorney direction. privilege never attached
The court found Heppner acted on his own initiative. his attorney never asked him to use Claude. Without that direction, the communications could not qualify as attorney-client privilege or work product. The attorney-as-agent framework required for privilege never formed
→ Marcella is provisioned and directed by the attorney, not the client acting independently. Every Marcella session originates within the firm's supervised workflow. Judge Rakoff specifically wrote that attorney-directed AI use "might arguably" be privileged. that is exactly Marcella's model
HEPPNER PROBLEM 3
No audit trail. no proof of supervised, competent AI use
Even if Heppner had used a better platform, there was no record of supervised, attorney-directed AI use that could demonstrate professional competence. ER 1.5(e) requires attorneys to use AI with supervision and verification. and to be able to prove it
→ Every Marcella session is linked to the named matter, timestamped, and auditable. The supervision record ER 1.5(e) requires is created automatically. not assembled after the fact when a bar complaint arrives
HEPPNER PROBLEM 4
Attorneys are not advising clients about AI risk. a new competence obligation
Heppner used public Claude on his own because no one told him not to. His attorneys did not know he was doing it. The court found this fatal to his privilege claim. Going forward, the duty of competence now includes advising clients about safe AI use. and providing a privileged alternative
→ Marcella gives attorneys a privileged AI framework to offer clients. one that is enterprise-secured, attorney-directed, and confidentiality-preserving. The advice "use Marcella, not public AI" is now a professional obligation. The tool that makes that advice actionable is here
The Daubert Two-State Model
Every Marcella output starts as AI GENERATED. Clearly labeled. Cannot be removed. The supervising attorney reviews, completes a six-item attestation checklist, and attests.
The output becomes ATTORNEY VERIFIED. The record is sealed and immutable. Outputs in AI GENERATED status cannot be added to matter files or used in attorney work product without first completing attorney review. The verification gate is enforced by the system. It is not optional and it cannot be bypassed.
Attestation checklist: source is real. citation exists and is retrievable. secondary treatment checked. jurisdiction confirmed. attorney reviewed. attorney attests output is appropriate.
U.S. v. Heppner, 25 Cr. 503 (JSR) · S.D.N.Y. · Feb. 17, 2026 · Judge Jed S. Rakoff · First impression nationwide
All Capabilities
What Marcella does —
and what that means for your practice.
01
Universal Legal Reference Layer. The Category No Competitor Has.
Westlaw owns the text of Williston, Prosser, Areeda, and the Restatements. Marcella owns something more useful: the empirical record of how courts in each jurisdiction have actually applied every section of those works. which sections they cite, how they treat them (followed, distinguished, limited, overruled), and how that treatment has shifted over time. An attorney researching a contract formation question in Washington doesn't just get "here is what Williston says." They get: "WA appellate courts have cited Williston § 4.8 in 23 opinions. followed in 19, distinguished in 4. here is the pattern of application and where WA diverges from TX." That is not a search result. That is a doctrine map. It does not exist anywhere else
● Live in production · Patent filed 2026
02
50-State Statutory Research. Answered in Plain English
Full statutory codes for all 50 states plus DC, refreshed monthly from official legislature feeds. Federal regulations via eCFR. CourtListener federal circuit and state appellate opinions. State AG opinions for all 50 states. Congressional Research Service reports. Ask in plain English. no Boolean syntax, no legal training required. The question that your associate spent three hours on. Marcella answers in 90 seconds with citations. You bill the same rate
● Live in production
03
Every document, query, and result is linked to a named matter. Sessions pick up exactly where they left off. next week, at deposition prep, at trial. Multiple team members contribute to the same matter thread. When you return to a matter six months later, Marcella already knows it. You don't re-brief. You don't re-research what you already researched. The value of every hour you've spent accumulates. That is not how any other tool works
● Live in production
04
Any legal document. contract, pleading, intake form, demand letter. classified with a risk level and the statutory or case law citation behind every flag. Your paralegal or junior associate submits documents for triage. A senior law pro reviews the flagged items, not the clean ones. Every contract your client sends is classified before it reaches a partner's desk. The work that used to require partner time gets done at associate cost. with a citation trail that satisfies professional responsibility review
● Live in production
05
Contract & Clause Analysis. 500+ Contract Types
CUAD-trained clause analysis across 500+ contract types. AIA A201 construction, pharma supply agreements, tech IP contracts, PI contingency fee agreements, government contracts (FAR/DFARS). Every flagged clause includes the industry-standard alternative and the legal authority behind the concern. Works inside Microsoft Word. research and redline suggestions without leaving the document you're drafting
● Live in production
06
eDiscovery Intelligence. Privilege Review at Fraction of Contract Attorney Cost
Bulk document review with GREEN / YELLOW / RED privilege classification. Custodian timeline reconstruction. Keyword-in-context analysis across large document sets. Legal holds tracked within the matter memory layer. every document touched, every classification decision logged, the complete audit trail built automatically. Privilege review that used to take a contract attorney three days takes an afternoon. The same corpus grounding that prevents hallucinated citations in research prevents misclassification in document review
● Live in production
07
Multi-Source Research Synthesis. Parallel Analysis With Confidence Signals
Parallel agents run simultaneously across statutory corpus, case law, AG opinions, treatise citation layer, and secondary authority. Synthesizes a structured answer with every source cited. The key discipline: when the law is unsettled. a circuit split, evolving doctrine, an open question. Marcella flags it. It does not fabricate an answer to an unresolved question. The confidence signal distinguishes what is settled from what is not. That is the standard a sophisticated practitioner requires and what general-purpose AI cannot provide
● Live in production
08
Ethics & Professional Responsibility Compliance
State bar ethics opinions indexed for all 50 states. ER 1.5(e) compliant. every output cites the indexed source, every session is linked to the named matter and timestamped. The supervision and verification record that bar ethics guidance requires for AI-assisted research is created automatically. Marcella's architecture makes the Heppner scenario. hallucinated citations. architecturally impossible. See the dedicated Heppner section above
● Live in production
09
Data Privacy. Client Data Never Leaves Your Microsoft Account
M365 tenant isolation. client data never crosses tenant boundaries. Microsoft Presidio PII anonymization strips identifying information before any content reaches the AI model. The AI reasons about anonymized derivatives. it never sees raw client names, matter IDs, or identifying details. Marcella does not train on your queries. Customer-managed encryption keys. Attorney-client privilege protected architecturally, not by policy. The same infrastructure as your firm's email. no new security review required
● Live in production
10
Paralegal & Staff Leverage. Plain English, No Training Required
Every capability accessible in plain English inside Teams and Word. No Boolean search syntax. No legal training required to get a useful answer. A paralegal or junior associate with no research background can triage documents, run statutory lookups, and compile research summaries. with the full citation trail and professional responsibility audit record attached. When a question exceeds scope, Marcella flags it for attorney review rather than fabricating an answer. The floor of your practice rises without adding headcount
● Live in production
12
1099-DA Crypto Tax Intelligence. Connect Wallet, Get Organized Output
Form 1099-DA reporting for digital assets is a new IRS requirement that most taxpayers. and many CPAs. are not prepared for. Marcella connects to wallet transaction history, maps transactions to the IRS's cost basis and gain/loss methodology, and produces an organized output that satisfies the 1099-DA reporting standard. No legal background required. a non-legal professional can activate Marcella, connect their wallet, and receive a structured tax document ready for their CPA or for self-filing. The time savings versus manual transaction-by-transaction reconciliation is substantial. The accuracy is architectural. Marcella applies the current IRS digital asset rules indexed in the corpus to the actual transaction record, not a best-guess interpretation
● Live in production
First-Mover Corpus · Clock Running
TXSE listings begin
in 2026.
The Texas Stock Exchange received SEC approval in September 2025. BlackRock, Citadel Securities, JPMorgan, Goldman Sachs, and Charles Schwab are founding investors. TXSE listing rules, corporate governance requirements, and TSSB enforcement posture are all new. No attorney has experience with them yet.
Marcella has the TXSE corpus indexed now. The attorney who has it today walks into every TXSE client call already knowing the answer. The attorney who waits reads the same PDF the client just read.
The attorneys who have the corpus today own the client relationships when listings start. This gap is not recoverable once it opens.
What Marcella Has Indexed
TXSE Listing Rules — full rulebook as filed with SEC
Texas Securities Act — full statute and TSSB enforcement actions
TSSB Board Rules — regulatory posture and guidance
SEC EDGAR — Delaware Court of Chancery, securities enforcement
PAT-04 treatise citation layer — securities law secondary authority mapped
Defense Contractors & Government Suppliers
government certification prep
assembles itself.
If your company holds government contracts involving sensitive government data, CMMC certification is a legal requirement. and the evidence package to prove it used to take months to compile manually. Marcella maps every policy, procedure, and security document to its required control at ingest. When the assessment comes, the package is already organized. No last-minute scramble. No compliance consultant billing $400/hour to sort folders
110
Controls Indexed
L2
CMMC Level
CMMC Module. What It Does
Maps your System Security Plan to all 110 NIST 800-171 Rev 3 controls at ingest. automatically, without manual tagging
Identifies control gaps before your DIBCAC assessment. not during it
Generates the complete evidence package organized by control, ready to hand over
DFARS 252.204-7021 flow-down analysis. identifies which subcontractors must comply
Davis-Bacon and FAR Part 36 federal construction compliance for government contractors
CEM continuous monitoring. behavioral entropy audit trail between assessment cycles
Common Questions
What attorneys
actually ask.
Q
Does Marcella replace attorneys?
No. and it is not designed to. Marcella is a legal research and document analysis platform. Every output includes citations and every session is linked to the named matter for attorney review. The attorney decides. Marcella makes attorneys faster and better-resourced. it does not make the judgment calls that define legal practice
Q
Isn't this just Copilot or ChatGPT in an M365 wrapper?
No. and the difference is the reason this exists. General-purpose AI tools like Copilot, ChatGPT, and Gemini are optimized for fluency. They will generate a plausible-sounding citation to a case that does not exist. Marcella's outputs are grounded in the indexed corpus. it can only cite what is indexed. Beyond citation safety, the 50-state statutory corpus, the treatise citation intelligence layer, matter memory, CMMC evidence vault, and eDiscovery classification do not exist in any general-purpose AI tool. See the full comparison at Marcella vs. Alternatives
Q
What happens to our client data?
Client data stays inside your Microsoft 365 tenant. Marcella does not transmit client data to external servers. Presidio PII anonymization strips identifying information before any content reaches the AI model. Marcella does not train on your queries. The data residency, security, and privacy obligations you already meet with Microsoft. Marcella operates within those same boundaries
Q
How current is the corpus?
Statutory codes are refreshed monthly via official legislature feeds with automated self-healing. CourtListener case law is continuously indexed. State AG opinions and CRS reports are updated on a regular refresh cycle. When a source changes, all dependent corpus entries are flagged for re-verification. you are not served stale authority without a confidence disclosure
Q
What happens when a client asks Marcella for legal advice?
Marcella is configured to recognize when a client inquiry crosses from information into legal advice — questions about legal strategy, predicted outcomes, or case-specific recommendations. When that boundary is reached, Marcella routes the question to the supervising attorney rather than answering it. The client receives a prompt to contact their attorney directly. The routing event is logged in the matter record. This is not a policy — it is enforced by the system. The attorney-client relationship requires that legal advice come from the attorney, not from a research and analysis platform. Marcella is built to preserve that boundary.
Q
How does Marcella prevent AI-generated citation errors?
Architecture, not policy. Marcella's AI retrieves from the indexed corpus. it does not generate legal authority. If a statute or case is not indexed, Marcella says so. It does not fabricate a plausible-sounding alternative. Every citation returned exists in the corpus. Every session includes the source reference. The complete Heppner protection architecture is described in the dedicated section above