Glassy User Guide
Version 3.0 | March 2026
Your AI assistant for California Workers' Compensation. The AI does the homework. You make the decisions.
| Product | URL |
|---|---|
| Glassy | glassy.adjudica.ai |
| Adjudica | app.adjudica.ai |
| These Docs | foundersdocs.adjudica.ai |
What Is Glassy?
Glassy is your personal AI assistant built specifically for California Workers' Compensation law. It handles conversations, research, analysis, and strategy — everything that happens between reading a document and making a decision about a case.
How it relates to Adjudica: Adjudica handles your documents and court forms. Glassy handles your conversations, research, and analysis. They are companion products, not competitors, and they work better together than apart.
The Glass Box promise: Most AI tools are a closed kitchen. You send in a question and get back an answer, with no way to see what happened in between. Glassy is a glass-walled kitchen. Every phase of AI reasoning is visible, inspectable, and auditable. You can see how it thinks, not just what it thinks. This is not a marketing distinction — it is an architectural one.
Security: Your data is encrypted with AES-256 at rest and TLS 1.3 in transit. Glassy is HIPAA compliant with a BAA available for Enterprise plans. Your conversations, memories, and case data are never used to train AI models. They are yours and they stay yours.
One more thing: The AI does not give legal advice. That is your job, and it always will be.
Quick Reference
Not a "read the whole manual" person? We get it. Here is the cheat sheet.
| I want to... | Where |
|---|---|
| Ask a question about a case | Chat in the sidebar |
| Calculate a PD rating | Skills > PD Calculation |
| Research case law | Skills > Case Research |
| Analyze a document | Skills > Document Analysis |
| Search legal knowledge | Knowledge in the sidebar |
| See how the AI thinks | Algorithm in the sidebar |
| Teach the AI about my practice | Practice Profile in the sidebar |
| Change how the AI talks to me | Personality in the sidebar |
| Save something for the AI to remember | Memory in the sidebar |
| Track work items | Tasks in the sidebar |
| Set up my case details | Cases in the sidebar |
| Manage my team | Organizations in the sidebar |
| Change my settings | Settings in the sidebar |
That is the whole platform in thirteen rows. Read on for details, or just start clicking — every sidebar icon has a tooltip if you hover over it.
1. Getting Started
Creating Your Account
Go to glassy.adjudica.ai and click Get Started. You can register with:
- Email and password — enter your name, email, and a password
- Google, Microsoft, or Apple — one click, no password to remember
If you go the password route, there are five requirements (shown as a live checklist while you type):
- At least 12 characters
- At least one uppercase letter
- At least one lowercase letter
- At least one number
- At least one special character
Yes, "Password1!" is technically valid. No, please do not use it.
First Login
The first time you open Chat, Glassy walks you through a brief onboarding conversation. It asks about your practice, your goals, and how you like to communicate. This is not busywork — your answers seed your Practice Profile and Memory, which means the AI starts giving you relevant responses from conversation one instead of conversation twenty.
You can skip the onboarding and build your profile manually later. But the five minutes you spend here pay dividends immediately.
Two-Factor Authentication
After you are set up, go to Settings > Security and enable two-factor authentication. It works with any TOTP app (Google Authenticator, Authy, 1Password). You scan a QR code, enter the six-digit verification, and save your backup codes somewhere secure. A password manager is ideal. The same device as your authenticator app is not.
The Dashboard
After logging in, you land on the Dashboard. It is organized into five categories of feature cards:
| Category | What is in it |
|---|---|
| Legal Tools | PD Calculator, Case Research, Document Analysis |
| Core | Chat, Skills Library, Task Board |
| Intelligence | Algorithm, Memory, Knowledge |
| Customization | Personality Studio, Practice Profile |
| Management | Organizations |
Each card is a quick-launch link to the corresponding feature page.
Sidebar Navigation
The left sidebar is your home base. It stays visible on every page:
| Icon | Feature |
|---|---|
| Home | Dashboard overview |
| Chat | AI conversations |
| Tasks | Task board |
| Skills | Structured AI tools |
| Knowledge | Semantic search |
| Algorithm | AI reasoning viewer |
| Memory | Context management |
| Practice Profile | Your professional identity |
| Personality | Communication style |
| Goals | Goal tracking |
| Organizations | Team management |
| Settings | Account and preferences |
Each icon has a tooltip. Hover if you forget what something does.
2. Chat
Chat is the main way you interact with Glassy. Type a question, get an answer. It sounds simple because it is.
Having a Conversation
Type your message in the input box at the bottom and press Enter to send. Responses stream in word by word, so you can start reading before Glassy finishes writing. Press Shift+Enter if you need a new line without sending.
The streaming is not just cosmetic — for long, detailed responses, it means you are reading the first paragraph while the fifth paragraph is still being generated. More efficient than staring at a loading spinner.
What to Ask
Glassy is built for California Workers' Comp. It knows the Labor Code, the AMA Guides, WCAB procedures, and the practical realities of running a WC practice. Some examples:
- "What are the current TD rate minimums and maximums for 2026?"
- "Draft a demand letter for a cumulative trauma claim"
- "Calculate the PD rating for a 15% WPI to the lumbar spine"
- "What does Escobedo say about apportionment?"
- "Prepare me for tomorrow's MSC on the Garcia matter"
- "What discovery should I propound for a psyche claim?"
- "Summarize the main defenses to an AOE/COE challenge"
You are not limited to Workers' Comp questions — Glassy can help with general tasks too — but WC is where it is sharpest.
Managing Threads
Your conversations are organized into threads in the left panel of the Chat view. Each thread is a separate, independent conversation.
- New thread: Click the compose icon to start a fresh conversation. Start new threads for new topics — do not try to cram everything into one.
- Rename: Double-click the thread title to rename it. Press Enter to save, Escape to cancel.
- Auto-title: Glassy names your thread based on the first message. If "New Chat" is not descriptive enough for you, it will not be descriptive enough six months from now either. Rename it.
- Delete: Click the trash icon next to any thread. This is permanent. No undo.
Voice
If you enable voice in Settings, you can speak your question and hear the response read aloud. Useful when you are reviewing a file and your hands are busy, or when you just want to think out loud.
Saving Messages to Memory
See a response you want Glassy to remember? Click Save to Memory on any message. It goes to the WARM tier of your Memory system (more on that in section 6). Next time a related topic comes up, Glassy draws on that saved context automatically.
How It Works (The Glass Box)
When you type a message, Glassy assembles context from several sources before responding: your Practice Profile (who you are and what you care about), your Memory (past conversations and preferences), your Personality settings (how you want the AI to communicate), and the Knowledge Base (legal principles and case law). Then it processes your request through a multi-phase thinking pipeline — you can watch this happen on the Algorithm page.
The result is a response shaped specifically for you, your practice, and your question.
Confidence scoring and the Algorithm audit trail let you verify the reasoning behind any response.
What this means for you: Every response is tailored to your practice, not generic boilerplate.
3. Skills Library
Skills are the core value proposition. While Chat is a freeform conversation, Skills are structured AI tools that do specific legal tasks with specific inputs and specific outputs.
Think of Chat as talking to an associate. Think of Skills as filling out an intake form and getting back a memo. Both are useful — Skills just give you more predictable, repeatable results when you need a specific deliverable.
Legal Skills
These are the main event. Twelve skills built for California Workers' Comp:
| Skill | What it does |
|---|---|
| Case Research | Searches WCAB decisions, case law, and regulations for arguments and precedent relevant to your issue |
| PD Calculation | Calculates permanent disability ratings using the AMA Guides — body parts, WPI, age, occupation, all the variables |
| Document Analysis | Reads medical reports, depositions, and other documents to extract key facts, dates, and conclusions |
| Settlement Analysis | Evaluates settlement proposals against case facts, ratings, and comparable outcomes |
| Hearing Prep | Organizes issues, evidence, and arguments for WCAB hearings into a structured preparation package |
| Deposition Prep | Generates question outlines for depositions based on the case context you provide |
| Medical Summary | Summarizes physician reports — diagnoses, treatment plans, restrictions, work status, and impairment findings |
| Timeline Builder | Creates a chronological case history from the facts you provide — dates, events, and medical milestones |
| Letter Drafter | Drafts demand letters, correspondence, and communications in your preferred style |
| Discovery Generator | Generates discovery requests tailored to your case — interrogatories, document demands, or requests for admission |
| Compliance Checker | Flags potential ethics or billing compliance issues based on the information you provide |
| Billing Tracker | Tracks billable activities and time entries for the work Glassy helps you with |
How to Use a Skill
- Go to Skills in the sidebar
- Find the skill you want (browse or search)
- Click it — a dialog opens with input fields specific to that skill
- Fill in the details. The more context you provide, the better the result. "Lumbar spine" gets you a generic answer. "15% WPI to the lumbar spine, 52-year-old warehouse worker, DOI 3/15/2023" gets you a useful one.
- Click Execute
- Review the output — edit, save, or put it to work
If you have an active Case set (see section 11), legal skills automatically pull in the case details. You do not need to re-enter the claimant's age and body parts every time you run a PD calculation. Set the case once, run skills as many times as you need.
General Skills
Glassy also includes general-purpose skills: Summarize, Email Draft, Meeting Summary, Task Prioritizer, Goal Breakdown, Daily Brief, and others. These are useful for practice management and general productivity, but the legal skills above are where the platform earns its keep.
How It Works (The Glass Box)
Skills are different from Chat. In Chat, you have a freeform conversation and the AI decides how to respond. With a Skill, you provide structured input (body parts, WPI percentage, age) and get structured output (a PD rating with full calculation breakdown).
The Skill reads your inputs, applies domain-specific logic, and produces a formatted result. If you have an active Case, the Skill also draws on those case details.
Every Skill execution is logged and auditable through the Algorithm page.
What this means for you: Use Skills when you need a specific deliverable. Use Chat when you need to think through a problem.
4. Knowledge System
The Knowledge system is Glassy's research library. It holds legal principles, case law, regulations, and any custom documents you have added — and it searches by meaning, not just keywords.
Three Domains
| Domain | What is in it |
|---|---|
| General | Broad knowledge base — general reference material |
| Legal | Case law, regulations, Labor Code sections, WCAB decisions, and legal principles |
| Custom | Documents and knowledge you have added yourself |
Searching
Go to Knowledge in the sidebar, select a domain, type your question, and click Search.
Here is the important part: this is not keyword search. Traditional search looks for exact words — search for "apportionment" and you get every document that contains that word, whether relevant or not. Glassy's Knowledge search understands what you mean. Search for "splitting liability between injuries" and it finds passages about apportionment even if they never use that exact phrase.
Results are ranked by relevance on a 0-100% scale. Higher scores mean a closer conceptual match. A score above 80% is a strong match. Below 50% is a loose relationship — the passage is related but may not be on point.
Each result card shows the source, the relevance score, and a content snippet so you can evaluate before clicking through.
Query History
Your past searches are saved in the sidebar. Click any previous query to re-run it instantly. Useful when you come back to a research thread after a few days and cannot remember exactly what you searched for. Which happens more often than any of us would like to admit.
How It Works (The Glass Box)
Traditional search looks for exact keywords. Glassy's Knowledge search converts your question and the knowledge base into mathematical representations of meaning (called embeddings), then finds the closest matches. This is why "splitting liability between injuries" finds passages about apportionment — the meanings are close even though the words are different.
Results are drawn from curated legal sources and any custom documents you have uploaded.
Relevance scores tell you how strong the match is — high scores mean high confidence, not certainty.
What this means for you: Search the way you think, not the way a filing system is organized.
5. How Your AI Thinks (The Algorithm)
This is the Glass Box in action. Most AI tools show you an input box and an output. Glassy shows you the seven steps in between.
"Most AI tools are a closed kitchen. You get the dish but never see the recipe. Glassy is a glass-walled kitchen — every phase of thinking is visible, inspectable, and auditable."
The 7 Phases
Every time Glassy processes a request, it moves through seven phases. Here is what each one does, in terms that do not require a computer science degree:
| Phase | What Glassy does | In legal terms |
|---|---|---|
| Review | Reads your request, gathers context from your Profile and Memory, and characterizes the full scope of what you are asking | Taking the client's call and pulling the file |
| Analyze | Identifies what matters — constraints, goals, key issues, and things that must not be missed | Spotting the legal issues |
| Strategize | Decides the best approach, defines what "success" looks like for this specific request | Planning your argument |
| Draft | Creates the response or work product according to the plan | Writing the brief |
| Apply | Runs any needed tools, searches, or calculations — this is where Skills and Knowledge get invoked | Pulling case law, running the numbers |
| Check | Verifies its own work against your requirements and constraints | Proofreading and cite-checking |
| Improve | Notes what worked and what did not, for future reference | Filing the work product and updating the matter notes |
Reading the Algorithm Page
Navigate to Algorithm in the sidebar to see your execution history. Each execution appears as a card you can expand to see phase-by-phase detail:
- Status: Pending, In Progress, Completed, or Failed. Each phase shows its own status so you can see exactly where the process is.
- Duration: How long each phase took. Useful for understanding where the AI spent its time.
- Constraints: Rules the AI extracted from your request — things like "must be under $50,000" or "only lumbar spine." These come from what you typed plus your Practice Profile.
- Success Criteria: How the AI defined "done" for this task before it started working. This is set during the Strategize phase, not after the fact.
- Drift Score: How well the AI stayed on track throughout the execution. Lower is better — 0 means perfect alignment with the success criteria. If the score is high, the AI wandered.
- Verification Report: The AI's self-check results from the Check phase. This is where you can see whether the AI caught its own mistakes before delivering the response.
Why This Matters
This is not a nice-to-have feature. It means you can see when the AI misunderstood your question (Review phase), missed a constraint (Analyze phase), chose the wrong approach (Strategize phase), or drifted off track (Check phase). You can catch problems at the source instead of staring at a bad answer wondering what went wrong.
Green does not mean correct. It means the AI thinks it is correct. The Algorithm page is where you verify that.
How It Works (The Glass Box)
The Algorithm page is Glassy's most distinctive feature. Every request you send is processed through a structured seven-phase pipeline, and every phase is logged with its inputs, outputs, timing, and status.
The pipeline draws on your Practice Profile, Memory, active Case, and the Knowledge Base to assemble context before any reasoning begins.
Constraints are extracted before the response is generated, success criteria are defined before work starts, and drift scores are measured throughout — not retrofitted after the fact.
What this means for you: Full transparency into AI reasoning, so you can trust the good answers and catch the bad ones early.
6. Memory System
Glassy remembers what matters to your practice. Not everything, and not forever — it uses a three-tier system, like a filing system for context.
Three Tiers
| Tier | Analogy | What it holds | When it is used |
|---|---|---|---|
| HOT | Your desk | Frequently accessed facts and preferences — things relevant to almost every conversation | Every conversation |
| WARM | Your filing cabinet | Moderately accessed — past insights, saved messages, useful context that comes up regularly | When relevant to the current topic |
| COLD | Off-site storage | Archived memories — older context, completed matters, historical reference | Only when specifically needed |
How Memories Get Created
Two ways:
- Automatically. As you chat, Glassy notices useful facts and saves them. If you mention that your client is a 52-year-old warehouse worker, Glassy stores that. Next time you ask about PD calculations, it remembers — you do not have to repeat yourself.
- Manually. Click Save to Memory on any chat message, or go to the Memory page and click Add Memory to create one from scratch.
Each memory has: content (the actual information), a tier (HOT, WARM, or COLD), tags (for filtering and organization), and an importance score (0-10, which determines retrieval priority when context space is limited).
Managing Memories
- Promote: Move a memory up one tier. COLD becomes WARM, WARM becomes HOT. Use this when something becomes more relevant to your current work.
- Demote: Move a memory down one tier. HOT becomes WARM, WARM becomes COLD. Use this when something becomes less relevant.
- Edit: Update the content, tags, or importance of any memory.
- Delete: Remove a memory permanently. Delete it and it is gone — from the database and from the AI's context.
Why It Matters
The more Glassy knows about your practice, the better it gets. A fresh Glassy account gives you competent but generic Workers' Comp responses. After a few weeks of conversation, with memories accumulating about your cases, your strategies, and your preferences, the responses are noticeably more relevant.
This is not magic — it is context. The same way a human associate gets better after working with you for six months, Glassy gets better as its memory of your practice grows.
How It Works (The Glass Box)
Every time you chat with Glassy, it checks your memories for relevant context. Memories move between tiers based on access frequency — frequently referenced memories stay HOT, older ones drift to COLD.
When context space is limited, Glassy prioritizes memories by tier (HOT first) and importance score (higher first).
You control what Glassy remembers. Delete a memory and it is gone — from every future conversation.
What this means for you: Glassy gets smarter about your practice over time, without you having to repeat yourself.
7. Practice Profile (TELOS)
Your Practice Profile is your professional identity for the AI. It tells Glassy who you are, what you care about, and how you approach your work. The more complete it is, the more relevant every response becomes.
Practice Profile is WHO YOU ARE. Personality (section 8) is HOW THE AI TALKS.
The 10 Component Types
| Type | What it captures | Example |
|---|---|---|
| Mission | Your firm's long-term purpose | "Maximize recovery for injured workers in the Inland Empire" |
| Goal | What you are working toward | "Build expertise in cumulative trauma cases" |
| Challenge | Current obstacles | "High volume of QME scheduling delays" |
| Strategy | Approaches you use | "Aggressive early discovery to force quick settlements" |
| Problem | Recurring issues | "Opposing counsel consistently delays depositions" |
| Narrative | Stories and experiences that shaped your practice | "Won a $2M case on a novel apportionment theory" |
| Belief | Core values | "Every injured worker deserves competent representation" |
| Frame | Mental models you use | "Think of every case as a story with a beginning, middle, and end" |
| Hard Lesson | Difficult experiences | "Lost a case by underestimating the AME's credibility" |
| Insight | Learned wisdom | "Always depose the treating physician before the QME" |
Where to Start
Start with Mission, Goal, and Challenge. These three give the AI the most immediately useful context. A Mission statement like "maximize recovery for injured workers" tells the AI to weight its suggestions toward aggressive advocacy. A Challenge about QME delays tells it to factor scheduling realities into hearing prep and timeline analysis.
The completeness ring at the top of the Practice Profile page shows how much of your profile is filled. You do not need 100% to see results — even one Mission and one Goal make a noticeable difference.
Managing Components
- Add: Click Add Component, select a type, enter a title and description, set a priority level (1-10).
- Edit: Click any component to update its content or priority.
- Archive: Click the archive icon to hide a component without deleting it. Archived components are not injected into the AI's context but are preserved for later. Use this for completed goals, resolved challenges, or strategies you have paused.
- Delete: Permanently remove an archived component. Cannot be undone.
How It Works (The Glass Box)
Every component you add to your Practice Profile is injected into the AI's context when it processes your requests. If your Mission says "maximize recovery for injured workers," the AI weights its suggestions toward aggressive advocacy. If your Challenge mentions QME delays, the AI factors that into hearing prep and timeline analysis.
The profile is not cosmetic — it directly shapes the AI's reasoning pipeline at the Review and Analyze phases.
Priority levels determine which components get included first when context space is limited.
What this means for you: Tell Glassy who you are and what you care about, and it tailors every response accordingly.
8. Personality Studio
The Personality Studio controls how Glassy communicates — not what it says, but how it says it.
Think of it as choosing between the associate who writes five-page memos and the one who gives you the answer in two sentences. Both are good at their job. You just want different things at different times.
The 12 Trait Sliders
Each slider runs from 0 to 100. Most people find values in the 30-70 range feel natural for everyday work.
| Trait | Low end (0) | High end (100) |
|---|---|---|
| Formality | Casual, conversational | Formal, structured |
| Detail | Concise, just the answer | Comprehensive, thorough |
| Initiative | Reserved, waits for direction | Proactive, suggests next steps |
| Reasoning | Analytical, shows the logic | Intuitive, gets to the conclusion |
| Tone | Serious, all business | Humorous, lighter touch |
| Directness | Direct, straight to the point | Diplomatic, softens the edges |
| Risk | Conservative, safe recommendations | Bold, pushes boundaries |
| Structure | Structured, lists and headings | Flexible, flowing prose |
| Complexity | Technical, precise language | Accessible, plain English |
| Confidence | Cautious, heavy on qualifiers | Confident, decisive statements |
| Innovation | Traditional, established approaches | Innovative, creative angles |
| Warmth | Impersonal, professional distance | Warm, personable |
Multiple Profiles
Create different profiles for different contexts:
- Client Emails: Formal, diplomatic, warm, accessible
- Internal Memos: Direct, concise, analytical, structured
- Brainstorming: Innovative, bold, proactive, flexible
Click Activate on any profile to make it the active one (indicated by a green dot). Only one profile is active at a time.
Live Preview
As you adjust the sliders, the right side of the editor shows a live preview of the system prompt Glassy will use. This is the actual instruction text, not a summary. Read it to make sure the sliders are producing the behavior you want. If the preview says "respond with detailed, formal analysis" and you wanted bullet points, move the Detail and Structure sliders.
9. Goals and Milestones
Goals let you track what you are working toward — whether that is a case strategy, a practice development objective, or a firm management target.
Creating a Goal
Go to Goals in the sidebar, click Add Goal, and enter:
- Title — what the goal is
- Description — why it matters and what success looks like
- Category — how to group it (case strategy, firm management, professional development, etc.)
- Target date — when you want it done
Status Tracking
Each goal can be: Active, Paused, Completed, or Cancelled. Filter chips at the top of the page let you narrow the view by status.
Milestones
Break large goals into discrete steps. Click into a goal and add milestones — each one gets a checkbox. As you complete milestones, the progress bar fills proportionally. Five milestones, three checked, 60% progress. The math is not complicated. The accountability is the point.
10. Tasks
A simple Kanban task board. Three columns: Pending, In Progress, Completed.
Creating Tasks
Click New Task and enter a title, description (optional), priority, and due date (optional). Priorities are color-coded:
| Priority | When to use it |
|---|---|
| Low | Nice-to-have, no deadline pressure |
| Medium | Standard work, normal course of business |
| High | Time-sensitive, should not be deferred |
| Critical | Blocks other work, needs immediate attention |
Managing the Board
Move tasks between columns by opening the card and updating the status. Due date warnings appear when deadlines approach.
This is not trying to replace your practice management system. It is a place to track the things Glassy helps you work on — the research tasks, the drafting projects, the follow-ups that come out of a chat session.
11. Cases
Cases in Glassy are lightweight profiles for AI context, not a replacement for your case management system. They give Glassy the facts it needs to make its legal skills useful.
Creating a Case
Go to Cases in the sidebar and click New Case. Fill in:
- Case name — the claimant's name or a short identifier
- Case number — the official WCAB or ADJ case number
- Injury date — used in date-of-injury calculations
- Body parts — affected body parts, used in PD assessments
- Claimant age — at the time of injury, affects benefit calculations
- Occupational group — the claimant's job category, relevant to disability ratings
Setting the Active Case
Click Set Active on any case card. Once a case is active, legal Skills automatically receive the case details — you do not need to re-enter the claimant's age and body parts every time you run a PD calculation or case research.
Only one case can be active at a time. If a skill gives you unexpected results, double-check that the right case is active. It is the WC equivalent of filing in the wrong matter folder.
Status options: Active, Closed, Pending, Review. When a case has a calculated PD rating, it displays on the case card.
12. Legal Research
The Legal Research section provides browsable legal principles organized by category. Each principle is presented as an expandable card showing:
- The principle name and description
- The relevant statute or regulation reference
- Related case links with citations, year, and relevance scores
This is a good quick reference when you need to check a legal standard or find the citation for a principle you already know. For deeper research, use the Knowledge system (section 4) or the Case Research skill (section 3).
13. Organizations
Organizations let you bring your firm or team into Glassy and manage usage together.
Setting Up
Go to Organizations in the sidebar, click Create Organization, and enter a name. The URL slug is generated automatically.
Inviting Members
Click Invite Member, enter an email address, and select a role:
| Role | What they can do |
|---|---|
| Owner | Full control — manage members, settings, billing, delete the organization |
| Admin | Invite and remove members, manage settings, view usage data |
| Member | Standard access to shared features and resources |
If the person you invite does not have a Glassy account yet, they will be prompted to create one when they accept.
Budget Management
Set a monthly AI spend limit for the organization. The usage progress bar shows current spend as a percentage of the ceiling:
- Green: Well within budget
- Yellow: Getting close
- Red: Near or at the limit
Administrators get alerts when the budget is approaching its limit, so they can adjust before work is interrupted.
14. Settings
Profile
Edit your display name, email address, and profile photo. The Save button highlights when you have unsaved changes.
AI Preferences
- Default model: Choose from Gemini 2.0 Flash, GPT-4, Claude, and others. The default model is used for new conversations unless a specific configuration overrides it. Changing your default does not affect conversations already in progress.
- Response style: Balanced, Concise, or Detailed. This sets a baseline — your Personality profile overrides it with more fine-grained control.
- Streaming responses: On by default. Turn it off if you prefer to see the complete response at once.
- Memory system: On by default. Turn it off if you do not want Glassy to automatically extract and store memories from your conversations.
Voice
- Enable voice: Toggle text-to-speech for AI responses
- Voice selection: Choose from 12 available voices
- Speed: Adjust playback speed (slower for note-taking, faster for quick review)
Notifications
Toggle notifications for task completions, agent activity, and weekly summaries. Choose between email alerts and in-app notifications.
Two-Factor Authentication
See section 1 for setup instructions. To disable, go to Settings > Security and click Disable 2FA. Confirm with your password.
Integrations
MerusCase OAuth setup is available here for firms that use MerusCase for case management. Connect the integration to sync case data between the two platforms.
Data Management
- Export my data: GDPR and CCPA compliant data export. Downloads everything Glassy knows about you.
- Delete account: Permanently removes your account, data, and all associated content. There is a 30-day grace period during which you can change your mind. After that, deletion is irreversible.
15. Billing
Plans
| Plan | Price | Highlights |
|---|---|---|
| Founding Member | $499/month | Locked-in rate for early adopters. Limited to 20 spots. Full platform access, priority support, direct founder access. |
| Retail | $1,999/month | Full platform access, dedicated support, custom instances available. |
| Enterprise | Custom pricing | HIPAA compliance with BAA, SSO/SAML authentication, dedicated infrastructure, custom SLA, and white-glove onboarding. |
The Founding Member rate is exactly what it sounds like — a thank-you price for the firms that bet on us early. When those 20 spots are gone, they are gone.
Usage Tracking
Your billing dashboard shows:
- Daily message count — how many messages you have sent and received
- Token budget — your monthly allocation and how much you have used
- Visual progress bars — at-a-glance budget status
Progress bars are color-coded: green (plenty of room), yellow (watch your pace), red (approaching the limit). You will never be cut off mid-conversation — the alerts are there to help you plan, not to interrupt your work.
Manage your subscription, payment methods, and invoice history from the billing page.
16. Automation and Intelligence Features
Task Coordination
The task board from section 10 is powered by an agent coordination system behind the scenes. When you create a complex task — or when Glassy breaks down a complex request — the platform can split it into subtasks, route each to a specialized agent, and coordinate the results.
- Dependencies: One task can depend on another (calculate PD before drafting the demand letter)
- Priority routing: Critical tasks get processed first
- Status tracking: See every subtask, its status, and its output
How It Works (The Glass Box)
When you ask Glassy to do something complex — like "prepare a complete settlement demand package" — it may break that into multiple steps: calculate PD, summarize medical records, draft the demand letter, check compliance. Each step can be handled by a specialized agent optimized for that task.
The task board shows you every step, its status, and its output. You can intervene at any point.
Dependencies are enforced automatically — downstream tasks wait until their prerequisites complete.
What this means for you: Complex work gets broken down and tracked, not lost in a single conversation thread.
Hooks: Event-Driven Automation
Hooks let you set up rules: "If X happens, automatically do Y."
Events you can watch for (12+ types): task completed, memory created, goal updated, skill executed, message sent, thread created, component added to Practice Profile, and more.
Actions you can trigger:
| Action | What it does |
|---|---|
| Notify | Sends you an in-app alert |
| Summarize | Asks the AI to generate a recap of the event |
| Invoke Skill | Automatically runs a skill, passing the event context as input |
| Webhook | Sends an HTTP POST to an external URL (for connecting to other tools) |
Example: "Every time a Medical Summary skill completes, automatically save the key findings to Memory." Set it up once, never think about it again.
Hooks support condition-based filtering (only trigger when specific conditions are met) and cooldowns (prevent the same hook from firing too often — important for high-frequency events like "message sent").
To create a hook: go to the Hook settings, click Create Hook, choose an event, choose an action, set your conditions and cooldowns, and activate it.
How It Works (The Glass Box)
Hooks watch for events in your Glassy workflow and react automatically. When you set up a hook, you define three things: what event to watch for, what conditions must be true, and what action to take.
Each hook execution is logged with the triggering event, the action taken, and whether it succeeded or failed.
Cooldown periods prevent runaway automation from consuming your budget.
What this means for you: Automate the predictable parts of your workflow so you can focus on the judgment-intensive parts.
The Cognitive Pipeline
This is the engine behind the Algorithm page (section 5). Each time Glassy processes a request, it creates a Cognitive Execution — a complete, inspectable record of its thinking process.
Every execution records:
- Constraints extracted from your request
- Success criteria defined before generating a response
- Drift scores measured throughout the process
- Artifacts generated with verification reports
- Phase-by-phase logs with timing, inputs, and outputs
Every execution is stored and reviewable. This is not a summary — it is the actual execution log.
"Your AI's thinking process, visible in real time — no black boxes, no trust-us-it-works."
How It Works (The Glass Box)
A Cognitive Execution is a complete audit trail of one AI reasoning session. It records what the AI understood (constraints), what it aimed for (success criteria), what it produced (artifacts), and how well it stayed on track (drift scores).
If the AI gives you a wrong answer, you can open the execution and see exactly where it went wrong: did it misunderstand the question? Miss a constraint? Drift off topic?
Every execution is cryptographically logged and immutable.
What this means for you: Every AI response comes with a full paper trail — the legal profession's kind of accountability.
A Note on OpenClaw
Glassy's backend includes a code generation engine called OpenClaw. This feature is not exposed in the legal product interface — it is not relevant to attorney workflows and is mentioned here only for completeness. If you see references to it in the platform documentation, you can safely ignore them.
Security and Compliance
This section contains no jokes. It is not that kind of section.
- Encryption: AES-256 at rest. TLS 1.3 in transit. Your data is encrypted when it is stored and encrypted when it moves.
- Data isolation: Your data is completely separate from other users and organizations. No cross-tenant access. No shared context.
- HIPAA compliance: Business Associate Agreement available for Enterprise plans. The platform is architected for HIPAA compliance from the ground up — it is not a bolt-on.
- Audit trail: Every action on the platform is logged with an immutable HMAC cryptographic signature. Audit logs are retained for six years. Each entry can be independently verified to confirm it has not been tampered with.
- Your data is never used to train AI models. Your conversations, memories, case data, and Practice Profile stay yours. They are not aggregated, anonymized, or fed back into any model training pipeline.
- Data residency: US-based servers on Google Cloud infrastructure.
- Data export: Full GDPR and CCPA compliant data export available in Settings at any time.
- Account deletion: 30-day grace period, then permanent and irreversible deletion of all data.
Common Questions
Is my data safe? Yes. AES-256 encryption at rest, TLS 1.3 in transit, HIPAA compliant architecture, and your data is never used to train AI models. See the Security section above for the full picture.
Can Glassy give legal advice? No. Glassy provides research, analysis, calculations, and drafts. Legal judgment is yours. Every output requires attorney review before it goes anywhere near a filing, a client, or an opposing party. That faded fax from 2003 is going to give the AI a headache too — it is a tool, not a partner of record.
What if the AI gets something wrong? It will, eventually. When it does: check the Algorithm page to see where the reasoning went wrong. Was the constraint extraction off? Did it drift during drafting? Override any output you disagree with. If you see a persistent pattern of errors, contact support — it may indicate a data or configuration issue worth investigating.
Does Glassy work with Adjudica? Yes. Adjudica handles your documents and court forms. Glassy handles your conversations, research, and analysis. They share the same platform infrastructure and complement each other. See Using Glassy and Adjudica Together for the details.
Can I use Glassy on my phone? Glassy is a web application accessible from any modern browser, including mobile browsers. The interface adapts to smaller screens. Go to glassy.adjudica.ai from your phone and log in normally.
What AI models does Glassy use? Gemini, GPT-4, and Claude — selected based on task type and optimized for different kinds of work. You can set your preferred default model in Settings. The platform handles model routing automatically unless you choose to override it.
Will my conversations be private? Yes. Only you and members of your organization can see your data. No one at Glass Box Solutions reads your conversations, and the data is not accessible to other Glassy users.
What happens if I cancel? Your data is retained for 30 days after cancellation. During that window, you can reactivate your account and everything will be exactly where you left it. After 30 days, all data is permanently and irreversibly deleted.
Support
| Need | Contact | Response |
|---|---|---|
| General support | support@adjudica.ai | 1 business day |
| Direct to founder | alex@adjudica.ai | Same day |
| Security or data concern | security@adjudica.ai | Immediate (24/7) |
Something broken? Something confusing? Something you wish existed? We want to hear it. This platform is built by people who listen to attorneys, not by people who think they know better than attorneys.
Cross-References
- For document management and court forms, see the Adjudica User Guide.
- New to the platform? See Getting Started with Adjudica.
- Using both products? See Using Glassy and Adjudica Together.
The AI does the homework. You make the decisions.
Glass Box Solutions, Inc.