⚖️ Legal AI

Claude Code for Lawyers:
Build a Legal AI OS in a Weekend

Brief drafting, contract review, research organization, and client communication templates — all running automatically every session, without any new infrastructure. Here is the exact setup.

📖 18 min read 🗓 Updated April 2026 ⚖️ For attorneys, paralegals & legal ops 🔧 Includes full CLAUDE.md template
Section 1

What Claude Code Is — and Why Lawyers Are Adopting It Fast

Claude Code is Anthropic's terminal-based AI assistant. Unlike chat-based AI tools, it reads your project files — including a CLAUDE.md file, a brain/ directory of knowledge, and a .claude/rules/ folder of behavioral directives — automatically at the start of every session. You get an assistant that already knows your practice areas, your preferred motion structure, your client communication tone, and your matter management workflow, without you re-explaining it each time.

Search volume for "AI for lawyers" has grown over 380% since mid-2024. According to the American Bar Association's 2025 Legal Technology Survey, 42% of attorneys at firms with 100+ lawyers now use a generative AI tool weekly, up from 11% two years prior. Solo and small-firm practitioners are the fastest-growing adopter segment — driven primarily by the economics: a solo practitioner who automates 90 minutes of daily drafting work across a 250-working-day year recovers 375 billable hours annually.

Most lawyers start with ChatGPT or a legal-specific SaaS. The friction they discover: these tools require re-explaining context every session, don't follow firm-specific formatting rules, and can't hold institutional knowledge across matters. Claude Code solves all three problems. With a properly structured Legal OS — the configuration files that Brainfile provides pre-built for attorneys — Claude reads your professional identity, your practice-specific rules, and your current matter context before it writes a single word.

380%
Growth in "AI for lawyers" search volume since 2024
42%
Big-firm attorneys using generative AI weekly (ABA 2025)
375 hrs
Billable hours recoverable per year at 90 min/day saved
0
Additional infrastructure required. Runs on your machine.

Why Claude Code specifically — and not a legal AI SaaS?

Legal-specific SaaS tools like Harvey, CoCounsel, and Spellbook are genuinely powerful but carry three tradeoffs that matter for many practitioners: (1) per-seat costs that scale linearly with your team, (2) a closed configuration — you cannot encode firm-specific procedural rules or matter management conventions — and (3) your work product flows through their servers under their data policies.

Claude Code with Brainfile runs entirely on your machine. The only outbound call is to Anthropic's API, which you control. Your CLAUDE.md, brain/ directory, and rules files are plain text files that live in your file system — reviewable, version-controllable, and portable. The configuration itself becomes institutional knowledge that compounds over time rather than disappearing when you cancel a subscription.

This guide covers the exact configuration — the CLAUDE.md templates, brain/ directory structure, and rules files — that legal professionals need to run Claude Code as a domain-specific legal assistant.

Section 2

5 Legal Use Cases — With CLAUDE.md Snippets for Each

Each use case below includes the CLAUDE.md directive that activates it. These snippets go into your project CLAUDE.md — Claude reads them automatically at session start, no prompting required.

01
Contract Review & Redlining
Automated risk-flagging, clause-by-clause analysis, and redline generation in your jurisdiction
CLAUDE.md — Contract Review Section markdown
## Contract Review Protocol ### Default jurisdiction: Illinois (adapt for other states) When reviewing any contract: 1. First, identify: contract type, governing law clause, parties 2. Flag ALL of the following with [RISK-HIGH], [RISK-MED], or [RISK-LOW]: - Unlimited indemnification clauses - One-sided termination rights - IP assignment language broader than work-for-hire - Limitation of liability caps below 1x contract value - Auto-renewal clauses with short opt-out windows - Jurisdiction/venue provisions outside client's home state - Non-compete scope exceeding 12 months or 50-mile radius (IL standard) 3. For each flagged clause, produce: [ISSUE], [WHY IT MATTERS], [SUGGESTED REDLINE] 4. End with executive summary: risk count by severity + 3 priority items Format output as: - Executive Summary (bullet points) - Clause-by-Clause Analysis (table: Clause | Risk Level | Issue | Redline) - Recommended Negotiation Priorities (numbered list)
02
Legal Research Organization
Structure, synthesize, and store case law research in a queryable brain/ directory
CLAUDE.md — Research Protocol markdown
## Legal Research Organization When I provide case citations, research notes, or Westlaw/Lexis exports: 1. Extract: case name, citation, court, year, holding, key dicta 2. Tag with practice area from: [EMPLOYMENT | CONTRACT | LITIGATION | CORPORATE | IP] 3. Write a 2-sentence brief (holding + why relevant to matter at hand) 4. Identify favorable vs. adverse authority — label clearly 5. Note any circuit splits, recent reversals, or pending cert Output format for saving to brain/research/: # [CASE NAME] — [CITATION] **Court:** | **Year:** | **Tags:** | **Favorability:** **Holding:** (1 sentence) **Relevance:** (1 sentence) **Key Quote:** (verbatim, with pinpoint) **Watch for:** (circuit conflicts, negative treatment) On session start, read brain/research/ and brief me on: - How many authorities per practice area - Any new adverse treatment I should know about
03
Brief Drafting
Motion practice, summary judgment briefs, and appellate arguments structured to your court's preferences
CLAUDE.md — Brief Drafting markdown
## Brief Drafting Standards ### Writing style - Persuasive, not academic. Every sentence advances a position. - Lead with the standard of review, then facts favorable to our client - Topic sentences state the legal conclusion — the paragraph proves it - Avoid passive voice except in procedural history - No string cites. One strong case per point, with parenthetical - Signal adverse authority before opposing counsel does; address it head-on ### Structure defaults Motion to Dismiss (12(b)(6)): I. Introduction (1 para, 3 sentences max) II. Statement of Facts (chronological, plaintiff's version) III. Standard of Review (Twombly/Iqbal, 7th Cir. gloss) IV. Argument (IRAC per section heading) V. Conclusion (relief requested, verbatim prayer language) Summary Judgment: Add: Statement of Undisputed Material Facts (numbered, cited to record) Add: Response to SOF if reply brief On any draft request: 1. Ask: motion type, court, judge, page limit, our client is [plaintiff/defendant] 2. Read brain/research/ for applicable authority before drafting 3. Flag every legal proposition that needs a citation with [CITE NEEDED] 4. Output word count at end. Flag if over 80% of page limit.
04
Client Communication Templates
Status updates, engagement letters, settlement summaries, and adverse decisions — in your voice
CLAUDE.md — Client Communications markdown
## Client Communication Standards ### Voice and tone - Direct, warm, and jargon-free. Clients are not lawyers. - Translate legal terms: "summary judgment" → "a request that the judge rule without trial" - Never use "pursuant to", "hereinafter", "aforementioned" in client-facing writing - Always tell the client: what happened, what it means for them, what comes next - Close every client email with a specific next step and timeline ### Template types I use regularly Status Update: [Date] | [Matter] | [What happened] | [What it means] | [Next step + deadline] Settlement Offer Summary: Pros/cons table → recommendation → client decision required Adverse Decision Letter: What happened | Appeal options + deadlines | My recommendation Engagement Letter: Scope | Fee structure | Billing cycle | Out-of-scope triggers | Termination When drafting client communications: 1. Read brain/matters/ for current matter context before drafting 2. Draft at 8th-grade reading level (Flesch-Kincaid target: 60-70) 3. Flag any legal conclusions that require attorney review before sending 4. Always include: [MATTER NAME] in subject line, billing code in footer
05
Matter Management & Deadlines
Active matter tracking, deadline calendaring, and daily briefings on what needs attention today
CLAUDE.md — Matter Management markdown
## Matter Management ### Session startup (MANDATORY — read first) On EVERY session start, read brain/matters/active_matters.json and output: 1. Deadlines in next 14 days (sorted ascending) with matter name + action required 2. Matters with no activity in 7+ days (flag for follow-up) 3. Opposing counsel response windows that have expired (flag immediately) ### Matter file structure Each matter lives in brain/matters/[MATTER_ID]/ status.md — current posture, stage, next milestone deadlines.json — all deadlines with rule basis (FRCP 12, local rule 5.3, etc.) parties.md — all parties, counsel, judges, mediators notes.md — chronological running notes (append only, never edit) Deadline math rules: - Federal deadlines: exclude weekends + federal holidays (FRCP 6(a)) - State deadlines: check brain/rules/[STATE]_deadline_rules.md - Always calculate from the TRIGGERING EVENT, not from today - If triggering event is disputed, flag it: [DEADLINE DISPUTE — CONFIRM] When I add a new deadline: 1. Ask: triggering event, deadline type, court/jurisdiction 2. Calculate the deadline with your rule basis shown 3. Update brain/matters/[MATTER_ID]/deadlines.json 4. Confirm: "Deadline set for [DATE] per [RULE]. Please verify."
Section 4

brain/ Directory Structure for Lawyers

The brain/ directory is your persistent knowledge base. Claude reads it every session — it is where institutional knowledge accumulates. Here is the recommended structure for a legal practice, including what belongs in each file. Read the full memory guide for how Claude accesses these files.

brain/ ├── matters/ # Active matter files │ ├── active_matters.json # Master list: matter ID, client code, stage, next deadline │ ├── matter-2024-031/ │ │ ├── status.md # Current posture, stage, open issues │ │ ├── deadlines.json # All deadlines: event, date, rule basis, status │ │ ├── parties.md # All parties, counsel contacts, judges, mediators │ │ └── notes.md # Running chronological notes (append-only) │ └── matter-2025-007/ │ └── ... (same structure) │ ├── research/ # Vetted case law and authority │ ├── employment/ # Employment law authority │ │ ├── wrongful_termination.md │ │ ├── discrimination.md │ │ └── non_compete.md │ ├── contracts/ │ │ ├── formation_enforceability.md │ │ └── remedies_damages.md │ └── litigation/ │ ├── pleading_standards.md # Twombly/Iqbal + 7th Cir. gloss │ └── summary_judgment.md │ ├── knowledge/ # Practice intelligence files │ ├── client_roster.md # Client codes + engagement types (no PII) │ ├── adverse_parties.md # All adverse parties ever — for conflict checks │ ├── opposing_counsel.md # Counsel intel: style, settlement posture, tendencies │ ├── judicial_profiles.md # Judge preferences: oral arg, briefing, settlement │ └── recent_developments.md # New cases + rule changes — update monthly │ ├── rules/ # Court-specific procedural rules │ ├── il_deadline_rules.md # Illinois state court deadline rules │ ├── ndil_local_rules.md # N.D. Ill. local rules summary │ └── federal_holidays.json # Current year federal holidays for FRCP 6(a) math │ ├── templates/ # Reusable document templates │ ├── engagement_letter.md │ ├── settlement_demand.md │ ├── status_update_email.md │ └── closing_letter.md │ └── session_notes.md # Running log of session decisions (append-only)

What makes this directory valuable over time

The brain/ directory compounds. On day one it contains your templates and rules. By month six, it contains your vetted research, detailed judicial profiles built from your actual courtroom observations, opposing counsel tendencies learned across multiple matters, and a complete conflict-check database. Claude reads all of it at every session start. The configuration becomes your institutional knowledge — portable, version-controlled, and yours.

This is the fundamental difference between a chat-based AI tool (stateless, re-explains everything every time) and a Claude Code Legal OS (stateful, accumulates knowledge, knows your practice). Learn more about structuring these files in the Claude Code memory guide.

Section 5

.claude/rules/ Files for Legal Work

Rules files in .claude/rules/ extend CLAUDE.md with domain-specific behavioral rules. They are loaded automatically alongside CLAUDE.md and apply to every session. Here are three rules files every legal practitioner should configure. See the complete rules guide for syntax and loading order.

1. citation-rules.md — Citation Quality Enforcement

.claude/rules/citation-rules.md markdown
# Citation Quality Rules ## The Core Rule NEVER include a legal citation unless you are certain it is real. If uncertain: write [CITE NEEDED — verify: topic]. Do not fabricate. ## Citation Format (Bluebook 21st) Cases: Party v. Party, Vol Reporter Page, Pinpoint (Court Year) Statutes: § section (Year if relevant) Regulations: Vol C.F.R. § section FRCP: Fed. R. Civ. P. ## (always abbreviated this way) FRAP: Fed. R. App. P. ## ## Citation integrity checks Before including any case citation: - Is this case from the correct jurisdiction? - Is this the most recent controlling authority (no later reversal)? - Is the pinpoint cite to the specific proposition, not the general holding? - Is the parenthetical accurate to what the case actually held? ## What to do if unsure Write: [CITE NEEDED — {description of proposition}] # Never: fabricate a plausible-looking citation # Always: flag for attorney to verify with Westlaw/Lexis ## Parenthetical rules - Parentheticals are mandatory for all non-binding authority - Use present tense: "holding that..." not "held that..." - One sentence. Verb + conclusion. No ellipsis.

2. drafting-voice.md — Writing Style Enforcement

.claude/rules/drafting-voice.md markdown
# Drafting Voice Rules ## The non-negotiable writing rules These apply to every single document this office produces. Structure: - Topic sentence states the legal conclusion - Body proves it (rule → application → conclusion) - One rule per paragraph. One case per rule. One pinpoint. - Transitions must carry meaning — never "furthermore" or "additionally" Tone for court filings: - Formal, precise, confident. No hedging. - Active voice 90%+ of the time - No rhetorical questions in argument sections - Short sentences in high-stakes points (under 20 words) Tone for client communications: - Conversational, specific, action-oriented - Every paragraph ends with: what client needs to know or do - No legal jargon without immediate plain-English translation ## Automatic word replacements If I write these, replace automatically: utilize → use commence → begin or start terminate → end prior to → before subsequent to → after in the event that → if at this point in time → now notwithstanding → despite ## Forbidden constructions Flag and require rewrite: - "It is respectfully submitted that" → just submit it - "As previously noted" → note it again or cut it - "The instant case" → "this case" - "The aforementioned" → name it again - Double negatives in argument sections

3. matter-hygiene.md — Matter File Consistency Rules

.claude/rules/matter-hygiene.md markdown
# Matter Hygiene Rules ## File naming conventions Matter IDs: [YEAR]-[3-digit sequence] — e.g., 2025-031 Documents: [MATTER_ID]_[DOCTYPE]_[YYYYMMDD].docx Pleadings: [MATTER_ID]_[MOTION_TYPE]_v[N].docx Research: brain/research/[PRACTICE_AREA]/[TOPIC].md ## What goes in notes.md vs status.md status.md — current snapshot only. Always overwrite: - Stage: [PRE-SUIT | FILED | DISCOVERY | MSJ | TRIAL | APPEAL | CLOSED] - Next milestone + deadline - Outstanding items (numbered) - Last updated: [DATE] notes.md — append-only chronological log. Never edit past entries: - Format: [DATE] | [AUTHOR] | [NOTE] - Examples: call summaries, strategy decisions, client instructions - Never delete entries — add [SUPERSEDED: see entry DATE] if needed ## Deadline entry format (deadlines.json) Every deadline entry must include: "event": what triggers or is triggered "deadline_date": YYYY-MM-DD "rule_basis": exact rule citation (FRCP 12(b), ILCS §, etc.) "calculated_from": triggering event + date "status": PENDING | MET | WAIVED | EXTENDED "flags": any disputes about calculation or tolling ## On every session end 1. Summarize what changed in any matter this session 2. Update status.md for any matters discussed 3. Append to notes.md for any strategy decisions made 4. Confirm: no new deadlines were discussed without being logged
🔗
How rules files load

Files in .claude/rules/ are automatically loaded by Claude Code alongside your CLAUDE.md. They stack — each file adds to the behavioral context, not overwrites it. Read the rules file guide for the full loading order and how to use rules files for different practice groups or matter types.

Section 6

Ethics & Client Confidentiality: What Every Attorney Needs to Know

Using AI in legal practice raises real professional responsibility questions. Most state bars have now issued guidance. Here is a clear-eyed summary of what the rules require and how to configure your Legal OS to stay compliant.

⚠️
Not legal ethics advice

This guide is informational. It does not constitute legal ethics advice and does not establish an attorney-client relationship. Your jurisdiction's specific rules and bar guidance control. Consult your state's formal ethics opinions on AI use before implementing any AI-assisted workflow for client matters.

The duty of competence and AI (ABA Model Rule 1.1)

ABA Comment 8 to Model Rule 1.1 requires lawyers to maintain competence with "relevant technology." By 2026, this is widely interpreted by bar ethics committees to include understanding how generative AI tools process data and what safeguards apply. The duty of competence now cuts both ways: using AI tools effectively is increasingly expected; using them carelessly is a disciplinary risk.

As of early 2026, over 35 state bars have issued formal guidance or ethics opinions on AI use in practice. The majority take the position that using AI for legal work is permissible under Rules 1.1, 1.3, and 5.3, subject to appropriate supervision and confidentiality protections. A small number have issued caution opinions flagging data routing concerns. Check your jurisdiction's most recent formal opinion.

Client confidentiality and data routing (ABA Model Rule 1.6)

Rule 1.6 prohibits disclosure of information relating to representation of a client. The question AI tools raise is: does sending a prompt containing client facts to an API constitute disclosure? Most ethics opinions land here: the analysis is consent-plus-safeguards. A few specific points that matter for Claude Code users:

Supervision of AI output (ABA Model Rule 5.3)

Rule 5.3 requires proper supervision of nonlawyer assistance. Most ethics guidance now extends this to AI tools. The practical implication: every AI-generated draft that goes out the door needs attorney review. This is not a burden that Claude Code increases — if anything, the Legal OS makes supervision faster by flagging [CITE NEEDED] markers, outputting QC summaries, and structuring drafts to match your review checklist rather than requiring a complete structural re-read.

Disclosure to clients

An increasing number of bar opinions require disclosure when AI is used substantially in preparing work product, unless the client has consented in the engagement agreement. The simplest approach: add a one-sentence disclosure to your engagement letter template. "I use AI-assisted drafting tools in the preparation of legal documents. All work product is reviewed and approved by the supervising attorney." This satisfies most current guidance and signals transparency to clients who increasingly expect it.

Unauthorized practice of law

Claude Code does not practice law and does not hold itself out as doing so. The Legal OS is configured specifically to flag legal conclusions for attorney review, to use probabilistic language ("strong argument," "risk of adverse ruling") rather than definitive legal predictions, and to require attorney sign-off before any client-facing communication is considered final. These are not cosmetic guardrails — they are professional responsibility features built into the system architecture.

Get the Legal OS — Pre-Built & Production-Ready

Complete CLAUDE.md, brain/ directory templates, three .claude/rules/ files, and matter management dashboards — configured for attorneys, ready in under 10 minutes.

Get Brainfile Pro — $149/mo Try Free Starter Kit
✓ Annual plan: $119/mo ✓ Enterprise: $299/mo ✓ Instant download ✓ No recurring SaaS overhead
Section 7

Before and After: 5 Legal Tasks With and Without Brainfile

The difference is not "AI vs. no AI" — it is a configured, context-aware Legal OS versus a stateless chat tool that re-asks for context every session. Here is what that looks like for the five most time-intensive legal tasks.

Task Without Brainfile Legal OS With Brainfile Legal OS
Contract Review Re-paste context + risk framework + jurisdiction each session. 10-15 min setup. Output varies by prompt quality. No clause-level flag tracking. Say "review this contract." Claude already knows your jurisdiction, risk thresholds, and flag format. Clause table + executive summary + redlines on first pass.
Brief Drafting Generic structure, generic voice. Manually specify court, standard of review, citation format every time. Cites are hallucination-prone. Re-read every sentence. CLAUDE.md carries court structure, voice rules, and citation format. [CITE NEEDED] flags every unverified cite. QC summary at end. Half the attorney review time.
Research Organization Research notes in a folder. Must re-brief Claude on what you've found. No synthesis. No adverse-authority flagging across matters. Knowledge resets every session. brain/research/ accumulates vetted authority. Claude reads it at session start and cites from it before suggesting new cases. Adverse authority mapped across matters.
Client Status Update Manually re-enter matter posture, client's background, and communication preferences per email. Output is professional but generic. No billing code in footer. Read brain/matters/[ID] → draft email in your voice with matter context, correct reading level, billing code, and next-step structure — without being asked.
Deadline Calculation Ask Claude to calculate. Hope it applies correct rule. No rule citation shown. Manual verification required. No logging to matter file. Calculate deadline → show rule basis (FRCP 6(a), local rule) → log to deadlines.json → flag if calculation is disputed → calendar conservatively by default.

Pricing comparison: Legal OS vs. legal AI SaaS

Tool Monthly Cost Configurable Data Control Knowledge Persistence
Harvey AI $3,000–8,000+/seat/yr Limited Vendor servers Platform-dependent
CoCounsel (Thomson Reuters) $400–600+/seat/mo Minimal Vendor servers Platform-dependent
Spellbook $99–249/seat/mo Templates only Vendor servers No
Claude Code (bare) API usage (~$20–80/mo) Full — if you build it Local + Anthropic API Only if you build brain/
Brainfile Pro + Claude Code $149/mo all-in ✓ Full ✓ Local ✓ Compounds daily
Section 8

FAQ: 8 Questions Lawyers Ask About Claude Code

Do I need to be technical to use Claude Code and Brainfile? +
No coding background is required. Claude Code runs in a terminal (Mac or Windows), which you open the same way you open any application. Brainfile Pro includes a setup guide that walks through the four installation steps — installing Claude Code, placing the brain/ directory, installing the CLAUDE.md, and installing the rules files — with exact commands to copy-paste. If you can manage a file folder structure on your computer, you can set this up. The typical attorney setup time is 15–30 minutes from download to first use.
Is it safe to use Claude Code for client work under my bar's ethics rules? +
The answer depends on your jurisdiction and how you configure the system. As of early 2026, the majority of state bars that have issued formal guidance permit AI use in legal practice subject to competence, supervision, and confidentiality safeguards. The Legal OS CLAUDE.md includes a Confidentiality Rules section that instructs Claude to use coded client references, never suggest uploading privileged documents to external services, and flag any actual client names found in brain/ files for review. You should still review your state bar's most recent formal ethics opinion on generative AI and consult the current Anthropic data processing addendum before using the tool for any actual client matter.
Will Claude Code hallucinate case citations? +
Large language models can and do generate plausible-looking but incorrect citations — this is a documented limitation that every attorney using any AI tool needs to understand. The Legal OS addresses this at the architecture level rather than through wishful prompting. The citation-rules.md file in .claude/rules/ instructs Claude to flag any citation it cannot verify with [CITE NEEDED] rather than fabricating a confident-looking but wrong citation. Every brief draft outputs a QC line with a count of cite flags before you receive the document. The workflow is: AI drafts → attorney verifies every [CITE NEEDED] flag against Westlaw or Lexis → attorney signs the document. This is not materially different from how a thoughtful supervising attorney would use first-year associate work product. Never file AI-drafted citations without independent verification.
How does Claude Code compare to legal-specific AI tools like Harvey or CoCounsel? +
Harvey, CoCounsel (Thomson Reuters), and Spellbook are purpose-built legal AI platforms with significant capabilities — particularly around deep integration with legal research databases like Westlaw and Lexis. They are strong choices for large firms with significant per-seat budget. The tradeoffs compared to Claude Code with Brainfile: (1) Cost — Harvey starts at several thousand dollars per seat annually; Brainfile Pro is $149/month covering your entire practice. (2) Configurability — you cannot encode your firm's specific procedural standards, matter management conventions, or communication voice into those platforms; with Claude Code you can encode everything. (3) Data routing — your work flows through those vendors' servers under their DPAs; with Claude Code your files stay local and only your prompts (not your documents) go to Anthropic's API. For solo practitioners and small firms, the economics and configurability of Claude Code + Brainfile typically win.
Can I use this with a paralegal or associate on my team? +
Yes. The Legal OS is a file-based configuration — it is a set of files (CLAUDE.md, brain/ directory, .claude/rules/) that any member of your team installs into their Claude Code project. Each person runs their own Claude Code instance; there is no shared cloud component except Anthropic's API. You can version-control the brain/ directory and CLAUDE.md in a shared git repository so that updates (new research, updated matter status, rule changes) sync across the team. Rule of thumb: any file in brain/ that contains matter-specific information should be controlled carefully; the core templates and rules files can be shared freely. The Brainfile Pro Enterprise tier ($299/month) includes team deployment guidance and firm-wide brain/ architecture templates.
What practice areas is this most useful for? +
The Legal OS delivers the most value in practice areas with high drafting volume — commercial litigation (motions, briefs, demand letters), transactional work (contract review, NDA redlining, M&A diligence checklists), employment law (EEOC responses, offer letters, separation agreements), and general corporate (board minutes, operating agreements, compliance memos). It is less valuable in court appearances, depositions, negotiations, and any work that is inherently live and conversational. One useful frame: if you can describe the task as "produce a draft document given these inputs," Claude Code can automate most of the first-pass drafting. If the task requires real-time judgment in a room with other people, it cannot.
How long does it take to see a return on the configuration investment? +
Most attorneys who complete the full Legal OS setup — including populating brain/research/ with vetted authority and completing brain/matters/ for active files — report meaningful time savings within the first week. A common pattern: the first day is setup (15–30 minutes), the second day is the first real test on a familiar task, and by day three or four the workflow feels natural. The compounding return comes over weeks and months: each session your brain/ directory grows, your judicial profiles get richer, your research accumulates. After 90 days of consistent use, attorneys typically report saving 60–90 minutes per workday — the equivalent of 250–375 billable hours annually at any billing rate.
What is Brainfile — is it a separate application? +
Brainfile is not a separate application you install. It is a pre-built, professionally validated configuration package for Claude Code — the CLAUDE.md, brain/ directory templates, .claude/rules/ files, and setup documentation, delivered for a specific profession (in this case, legal practice). You download the Brainfile Legal OS package, place the files into your Claude Code project directory, customize your name, jurisdiction, and practice areas, and run Claude Code. From that point forward, Claude Code behaves as a configured legal assistant rather than a general-purpose AI. No additional software, no SaaS login, no recurring infrastructure to maintain. The free starter kit includes three templates so you can evaluate the approach before subscribing.

Your Law Practice,
Powered by Claude Code

Stop re-explaining your practice to an AI every session. The Legal OS gives Claude everything it needs to know about how you practice, what you value, and how your matters are structured — on day one.

Get Brainfile Pro — $149/mo Try Free Starter Kit
✓ Annual: $119/mo ✓ Enterprise: $299/mo ✓ No SaaS infrastructure ✓ Works with any Claude plan

Continue reading

CLAUDE.md Templates → Memory Guide → System Prompts Guide → Rules Guide →