The drafting and research desk built for Indian litigation.
Ingests every document in the matter. Cross-references your evidence with binding precedent. Drafts court-ready filings that cite the right era's statute — before you ask.
Encrypted end-to-end · No model training on your content · DPDP §6 aligned
Supreme Court + 25 High Courts · BNS/BNSS/BSA + IPC/CrPC/IEA era-aware · DPDP §6 aligned
"Does our affidavit cover the 7-step Bhajan Lal threshold?"
Your affidavit (Ex-A.pdf p.12) covers 5 of 7 grounds. Categories 4 (mala fide) and 6 (absurd inherent improbability) are missing — both surface in Bhajan Lal [1] and R.P. Kapur [2].
High Courts indexed
Judgments searchable
Draft templates
Format-review checks
Document IntelligenceNEW
A strategy memo for every matter — written before you ask.
Drop a folder of FIRs, statements, contracts, and exhibits into a Smart Folder. Lexarna reads everything, extracts parties + timeline + relief sought, runs first-pass contradiction detection across the file set, and synthesises a one-paragraph case theory plus an attack-and-defence memo. Open the Intelligence tab the next morning — it's ready.
"Plaintiff's title rests on a 2019 sale deed; defendant claims adverse possession from 1998. Strongest hook: defendant's own 2015 lease application names plaintiff's father as landlord — fatal to AP claim."
STRONGEST
2015 lease app — admission against interest
WEAKEST
No mutation entry 2019–2024
Research
Answers that quote your evidence and binding precedent — in the same reply.
Scope a research session to a matter and Lexarna ranks chunks from your case file alongside Supreme Court + High Court precedent in a single answer. Lexarna cross-compares them: where your evidence aligns with binding authority, where it diverges, and what's missing. Switch to MAX mode for agentic decomposition — the question is split into 4–5 sub-angles and answered in parallel.
QUERY
"What's our exposure on the Section 138 notice?"
Notice demands ₹4.2L — same as cheque amount. Section 138 satisfied.
Demand exceeding cheque amount renders the notice invalid — strict construction of S.138.
Divergence: Notice is clean but doesn't record dispatch method — defendable in cross.
Knowledge Graph
Find the contradiction before opposing counsel does.
Lexarna builds a knowledge graph of every person, event, and exhibit across the matter and runs two-layer contradiction detection: a first-pass that flags candidate pairs on every ingest, and an on-demand LLM verifier when the lawyer wants confirmation. Evidence is weighted by legal reliability so you know exactly how strong each contradiction is.
PW-2 places the incident at 9:45 PM; PW-3 records 10:30 PM.
Chargesheet names Sector 12 market; FIR places it at Sector 14.
Library
Two kinds of folders. One library.
Smart folders run the full intelligence pipeline — entity extraction, contradiction detection, strategy memo, cross-corpus research. Default folders are plain document storage for things that don't warrant the pipeline. You pick when you create the folder; you can promote a default folder to smart any time.
Sharma v. Mehta
Smart · 18 docs · 4 contradictions · memo ready
Acme Corp. — IPO due diligence
Smart · 42 docs · synthesising memo…
Bar Council CLE — 2026-27
Default · 6 docs · storage only
Templates — Bombay HC
Default · 12 docs · storage only
Drafting
112 templates covering every common Indian draft type.
Writs, criminal (bail/quashing/S.138), family law, MACT, DRT/CAT/AFT/NGT, RERA, IBC, DV Act, commercial contracts (MSA/SOW/DPA), tax/GST, and corporate filings. Before generation runs, an applicable-law preview shows exactly which sections will be cited — confirm or edit. Every draft then passes a three-layer format review, so what lands is court-ready, not a first pass.
Draft preview: In the Court of Sessions at New Delhi, Bail Application in the matter of Rajesh Kumar versus State (NCT of Delhi), arising out of FIR No. 234/2026, PS Karol Bagh, under section 318 BNS.
Find your practice area
Draft quality
Court-format guaranteed — three layers before you see it.
Every draft passes 43+ checks across three layers (33 deterministic + 10 LLM-judge). Layer 1 runs 33 deterministic checks — cause-title shape, GROUNDS lettered A./B./C., a canonical prayer opener, three-bucket verification. Layer 2 is 10 LLM-judge checks on semantics — party-label drift, era-mismatched statutes, fabricated citations, hallucinated headings. Layer 3 is a surgical auto-rewrite that fixes only the flagged spans and leaves the rest verbatim. Backed by 20+ practice-area format specs — ~165k words of Indian drafting rules.
Format review pipeline. Layer 1: 33 deterministic checks pass. Layer 2: catches a CrPC citation used for a post-July-2024 matter. Layer 3: auto-rewrite substitutes S.483 BNSS. Result: 0 violations remaining.
Self-learning
Drafts that improve every time you edit them.
Seed the Style Library with a few past drafts and Lexarna mirrors your voice from day one — auto-classifying each exemplar to one of the 112 template types. It then learns silently from your OnlyOffice edits and from your revise instructions, so citation conventions, prayer-clause length, and recurring arguments converge on your firm's house style. Memory is scoped per user, per org, and per matter — team consistency without cross-firm leakage.
Draft 1 uses a generic voice. After 10 revise instructions and 4 OnlyOffice edits, Draft 10 adopts the firm's voice — a 3-line prayer, lettered reliefs, and the firm's citation style.
Verification
Every answer passes 5 checks before you see it.
Generic AI tools hallucinate case names and confuse BNS with IPC. Lexarna runs every research answer through this verification pipeline before you see it — and every draft through the three-layer format review above.
Citation Grounding
Every [N] reference is validated against the retrieved corpus. Hallucinated case names are blocked before they reach the UI.
Factual Consistency
Claims in the answer are cross-checked against the cited excerpts. Fabricated entities and self-contradictions are caught.
Jurisdiction & Era
Rule-based check that the answer cites BNS/BNSS for post-July-2024 matters and IPC/CrPC for legacy ones. No mixed-era mistakes.
Legal Language
Enforces Indian legal terminology, honorifics, date formats, and section conventions. Reads like a lawyer wrote it.
Elaboration
Thin answers are detected and automatically expanded with supporting evidence — no one-sentence responses to complex questions.
Plus click-through citations. Every inline [N] opens the full judgment with the exact passage highlighted — court, bench, date, paragraph. Verify everything.
Thinking modes
Pick how deep to think.
From fast lookups to agentic decomposition. The credit cost is shown upfront — no surprises.
Quick
Fast tier
Fast lookups, known citations
1–3 credits
Standard
Standard tier
Balanced research and drafting
3–5 credits
Advanced
Advanced tier
Multi-step legal reasoning
15–25 credits
MAX
MAX · agentic
Decomposes into 4–5 sub-angles, retrieves for each in parallel
50–80 credits
Firm Memory
Insights from one matter help every matter.
Lexarna learns from your research turns — preferred citation style, recurring counter-arguments, judge-specific framing — and surfaces them on future questions when relevant. Scoped per user, per matter, and per firm. Auditable and revocable from Settings → Privacy.
"Justice Khanna prefers concise prayer clauses — 3-line max. Captured from a matter your firm worked on in March."
SCOPE
Your client data stays yours.
Documents and matters live in your private workspace, encrypted at rest, scoped to your organisation. We never train models on your case content.
A small set of opt-in toggles let you contribute anonymised search patterns or quality-check samples — all granular, all reversible. Full DPDP posture →
Questions
Try Lexarna with your next matter.
Free during private beta · Plans announced at launch
We're onboarding practitioners one at a time. Join the waitlist and we'll reach out when your spot opens.
Private beta · No credit card required · Existing users sign in