{"version":"1.0","updated":"2026-06-17T13:06:28.611Z","total":38,"pages":[{"slug":"admin","title":"SuperAdmin Guide","description":"Platform-wide administration for AI Assess Tech staff including user management, billing, question banks, and Ethereum anchoring","category":"product-guides","humanUrl":"/docs/admin","updated":"2026-03-02T00:00:00.000Z","content":"# SuperAdmin Guide\n\nPlatform-wide administration for AI Assess Tech staff. SuperAdmin features are only visible to users with the `SUPERADMIN` platform role.\n\n## System Dashboard\n\nReal-time overview of platform health and activity (SuperAdmin → Admin Portal):\n\n- **Total Users** — All registered users across all organizations\n- **Organizations** — Active organizations on the platform\n- **Total Tests** — Cumulative assessments since platform launch\n- **Tests Today** — Assessments in the last 24 hours\n- **Completion Rate** — Percentage of tests that completed successfully\n\n> A low completion rate (<80%) may indicate API issues, user drop-off, or timeout problems.\n\n## User Management\n\nManage all platform users (SuperAdmin → Users). Each user shows: name, email, organization, subscription tier, free credits remaining, organization role (MEMBER/ADMIN), platform role (blank or SUPERADMIN), and status (Active/Disabled).\n\nActions: Edit Role, Grant SUPERADMIN, Disable/Enable account, Add bonus credits.\n\n## Organization Management\n\nView and manage all organizations (SuperAdmin → Organizations). Each org shows: name, slug, user count, test count, and creation date. Click \"Review\" for detailed info including users, subscription tier, API key usage, and assessment history.\n\n## Billing Overview\n\nPlatform-wide revenue and subscription management (SuperAdmin → Billing Overview):\n\n- **Monthly Recurring Revenue (MRR)** — Total monthly revenue from active subscriptions\n- **Total Subscriptions** — Count of paying customers\n- **Total Overages** — Overage charges collected this period\n- **Past Due** — Subscriptions with failed payments (review immediately if > 0)\n\n## Lead Management\n\nTrack leads from AIAssessmentTool.com demo signups (SuperAdmin → Leads). Pipeline stages: Total → New → Contacted → Qualified → Converted → Won / Lost.\n\n> Leads with high assessment counts (10+) and recent activity are most likely to convert.\n\n## Question Banks\n\nManage cryptographically-locked question banks (SuperAdmin → Question Banks → All Banks). Each bank shows: framework, bank type, version, status, question count, and lock date.\n\n### Question Bank Lifecycle\n\n```\nDRAFT → SUBMITTED → (Review) → PUBLISHED → LOCKED → ANCHORED (Ethereum)\n```\n\n## Ethereum Blockchain\n\nManage blockchain anchoring for tamper-proof question bank verification (SuperAdmin → Ethereum → Overview).\n\nAI Assess Tech uses Ethereum mainnet for immutable, publicly verifiable proof of question bank integrity.\n\n- **Banks Anchored** — Question banks with on-chain proofs\n- **Ready to Anchor** — Locked banks awaiting blockchain anchoring\n- **Transactions** — Total Ethereum transactions executed\n\nActions: Anchor Bank (~$2–5 gas cost), Verify On-Chain, View Transaction History.\n\n## Project Planning\n\nInternal roadmap and milestone tracking (SuperAdmin → Plan):\n\n- Revenue Gate — First paying customer milestone\n- Investor Ready — Platform ready for fundraising\n- PRD Complete — Product requirements finalized\n- Patent Deadline — IP protection timeline\n\n### Next Generation Engines\n\n- **VIRTUE Engine (v2.0)** — Virtue-based assessment\n- **ETHICS Engine (v2.1)** — Deep ethical reasoning\n- **Operational Excellence (v2.2)** — Industry-specific compliance"},{"slug":"api-keys-anthropic","title":"Anthropic API Key Setup","description":"Step-by-step guide for obtaining an Anthropic API key for Claude 3.5 Sonnet, Claude 3 Opus, and Claude 3 Haiku","category":"get-started","humanUrl":"/docs/api-keys/anthropic","updated":"2026-03-02T00:00:00.000Z","content":"# How to Get an Anthropic API Key\n\nSetup time: 2–3 minutes. $5 free credits. Credit card required.\n\n## Before You Start\n\n- A valid email address\n- Phone number for verification\n- Credit card for account verification\n\n> Anthropic requires a credit card to create an API account, but new accounts receive $5 in free credits. You won't be charged until you exceed these credits.\n\n## Step-by-Step Instructions\n\n### Step 1: Go to Anthropic Console\n\nVisit [console.anthropic.com/signup](https://console.anthropic.com/signup).\n\n> The API console (console.anthropic.com) is different from Claude.ai. You need an API account to use Claude programmatically.\n\n### Step 2: Create Your Account\n\nSign up with Email or Google. Verify your email.\n\n### Step 3: Verify Your Phone Number\n\nAnthropic requires phone verification for security.\n\n### Step 4: Add Payment Method\n\nGo to Settings → Billing → \"Add payment method.\" New accounts receive $5 in free API credits.\n\n### Step 5: Generate Your API Key\n\nGo to Settings → API Keys or [console.anthropic.com/settings/keys](https://console.anthropic.com/settings/keys). Click \"Create Key,\" name it, and click \"Create Key.\"\n\n> Your API key is only shown once. Copy it immediately and store it securely.\n\n### Step 6: Copy Your API Key\n\nYour key looks like: `sk-ant-api03-aBcDeFgHiJkLmNoPqRsTuVwXyZ...`\n\n## Using Your Key with AI Assess Tech\n\n**Demo Tool:** Go to [aiassessmenttool.com](https://aiassessmenttool.com), select Anthropic, paste your key, choose Claude 3 Haiku (cost-effective), click \"Start Assessment.\"\n\n**SDK Integration:**\n\n```typescript\nimport Anthropic from '@anthropic-ai/sdk';\nimport { AIAssessClient } from '@aiassesstech/sdk';\n\nconst anthropic = new Anthropic({ apiKey: process.env.ANTHROPIC_API_KEY });\nconst client = new AIAssessClient({ healthCheckKey: process.env.AIASSESS_KEY });\n\nconst result = await client.assess(async (question) => {\n const response = await anthropic.messages.create({\n model: 'claude-haiku-4-5',\n max_tokens: 100,\n messages: [{ role: 'user', content: question }]\n });\n return response.content[0].type === 'text' ? response.content[0].text : '';\n});\n```\n\n## Pricing (as of 2026)\n\n| Model | Input (per 1M tokens) | Output (per 1M tokens) | Est. Cost per Assessment |\n|-------|----------------------|----------------------|------------------------|\n| Claude 3 Haiku (Recommended) | $0.25 | $1.25 | ~$0.02 |\n| Claude 3.5 Sonnet | $3.00 | $15.00 | ~$0.25 |\n| Claude 3 Opus | $15.00 | $75.00 | ~$1.20 |\n\n## Troubleshooting\n\n- **\"Invalid API Key\"** — Ensure the full key including \"sk-ant-\" prefix is copied. Check for extra spaces.\n- **\"Rate limit exceeded\"** — Anthropic has tier-based rate limits. Wait a minute and retry.\n- **\"Credit balance too low\"** — Add more credits at [console.anthropic.com/settings/billing](https://console.anthropic.com/settings/billing)."},{"slug":"api-keys-gemini","title":"Google Gemini API Key Setup","description":"Step-by-step guide for obtaining a Google Gemini API key for Gemini Pro, Gemini 1.5, and Gemini 2.0 Flash","category":"get-started","humanUrl":"/docs/api-keys/gemini","updated":"2026-03-02T00:00:00.000Z","content":"# How to Get a Google Gemini API Key\n\nSetup time: 3–5 minutes. Free tier included. No credit card required.\n\n## Free Tier Benefits\n\n- 15 requests per minute for Gemini 1.5 Flash\n- 1,500 requests per day included\n- No credit card required\n\nThis is enough for dozens of AI ethical assessments without any cost.\n\n## Before You Start\n\n- A Google account (Gmail, Google Workspace, etc.)\n\n## Step-by-Step Instructions\n\n### Step 1: Go to Google AI Studio\n\nVisit [aistudio.google.com/apikey](https://aistudio.google.com/apikey).\n\n### Step 2: Sign In with Google\n\nClick \"Sign in with Google\" and select your account.\n\n### Step 3: Accept Terms of Service\n\nReview and accept Google's AI Studio Terms of Service.\n\n### Step 4: Create Your API Key\n\nClick \"Create API key\" → \"Create API key in new project\" (or select an existing GCP project). If you don't have a Google Cloud project, AI Studio will create one automatically (free).\n\n### Step 5: Copy Your API Key\n\nYour key looks like: `AIzaSyAaBbCcDdEeFfGgHhIiJjKkLlMmNnOoPp`\n\n> While you can view this key again in AI Studio, store it securely right away.\n\n## Using Your Key with AI Assess Tech\n\n**Demo Tool:** Go to [aiassessmenttool.com](https://aiassessmenttool.com), select Google Gemini, paste your key, choose Gemini 1.5 Flash, click \"Start Assessment.\"\n\n**SDK Integration:**\n\n```typescript\nimport { GoogleGenerativeAI } from '@google/generative-ai';\nimport { AIAssessClient } from '@aiassesstech/sdk';\n\nconst genAI = new GoogleGenerativeAI(process.env.GEMINI_API_KEY!);\nconst model = genAI.getGenerativeModel({ model: 'gemini-1.5-flash' });\nconst client = new AIAssessClient({ healthCheckKey: process.env.AIASSESS_KEY });\n\nconst result = await client.assess(async (question) => {\n const response = await model.generateContent(question);\n return response.response.text();\n});\n```\n\n## Pricing (as of 2026)\n\n| Model | Free Tier | Paid (per 1M tokens) | Est. Cost per Assessment |\n|-------|-----------|---------------------|------------------------|\n| Gemini 1.5 Flash (Recommended) | 15 RPM, 1M tokens/min | $0.075 input / $0.30 output | Free* |\n| Gemini 1.5 Pro | 2 RPM, 32K tokens/min | $1.25 input / $5.00 output | ~$0.08 |\n| Gemini 2.0 Flash | 10 RPM | $0.10 input / $0.40 output | ~$0.01 |\n\n*Within free tier limits. Check [ai.google.dev/pricing](https://ai.google.dev/pricing) for current rates.\n\n## Troubleshooting\n\n- **\"API key not valid\"** — Ensure the full key is copied. Gemini keys start with \"AIza\". Verify the correct API endpoint.\n- **\"Quota exceeded\"** — Free tier rate limits hit. Wait a minute for RPM limits, or next day for daily limits. Enable billing for higher limits.\n- **\"Permission denied\"** — Enable the Generative Language API in your Google Cloud project at [Google Cloud Console](https://console.cloud.google.com/apis/library/generativelanguage.googleapis.com)."},{"slug":"api-keys-grok","title":"xAI Grok API Key Setup","description":"Step-by-step guide for obtaining an xAI Grok API key for Grok 2 and Grok Beta with $25 free credits","category":"get-started","humanUrl":"/docs/api-keys/grok","updated":"2026-03-02T00:00:00.000Z","content":"# How to Get an xAI Grok API Key\n\nSetup time: 2–3 minutes. $25 free credits. No credit card required initially.\n\n## $25 Free Credits\n\nxAI offers generous free credits: $25 in free API credits upon signup, no credit card required, credits don't expire (unlike some providers). This is enough for hundreds of AI ethical assessments.\n\n## Before You Start\n\n- A valid email address\n- Phone number for verification (may be required)\n\n> xAI is the company behind Grok, founded by Elon Musk. The API is separate from the Grok chatbot on X (Twitter).\n\n## Step-by-Step Instructions\n\n### Step 1: Go to xAI Console\n\nVisit [console.x.ai](https://console.x.ai).\n\n### Step 2: Create Your Account\n\nSign up with Email, X (Twitter), or Google.\n\n### Step 3: Verify Your Identity\n\nxAI may ask for phone verification depending on region and account type.\n\n### Step 4: Navigate to API Keys\n\nGo to the API Keys section or directly to [console.x.ai/api-keys](https://console.x.ai/api-keys).\n\n### Step 5: Create Your API Key\n\nClick \"Create API Key\" or \"+ New Key,\" name it, and click \"Create.\"\n\n> Your API key is only shown once. Copy it immediately and store it securely.\n\n### Step 6: Copy Your API Key\n\nYour key looks like: `xai-aBcDeFgHiJkLmNoPqRsTuVwXyZ123456789...`\n\n## Using Your Key with AI Assess Tech\n\n**Demo Tool:** Go to [aiassessmenttool.com](https://aiassessmenttool.com), select xAI Grok, paste your key, choose Grok 2, click \"Start Assessment.\"\n\n**SDK Integration:**\n\n```typescript\nimport { AIAssessClient } from '@aiassesstech/sdk';\n\nconst client = new AIAssessClient({ healthCheckKey: process.env.AIASSESS_KEY });\n\nconst result = await client.assess(async (question) => {\n const response = await fetch('https://api.x.ai/v1/chat/completions', {\n method: 'POST',\n headers: {\n 'Authorization': `Bearer ${process.env.XAI_API_KEY}`,\n 'Content-Type': 'application/json'\n },\n body: JSON.stringify({\n model: 'grok-2',\n messages: [{ role: 'user', content: question }]\n })\n });\n const data = await response.json();\n return data.choices[0].message.content;\n});\n```\n\n> Grok's API is compatible with the OpenAI SDK format. You can use the OpenAI SDK by changing the base URL to `https://api.x.ai/v1`.\n\n## Pricing (as of 2026)\n\n| Model | Input (per 1M tokens) | Output (per 1M tokens) | Est. Cost per Assessment |\n|-------|----------------------|----------------------|------------------------|\n| Grok 2 (Recommended) | $2.00 | $10.00 | ~$0.15 |\n| Grok 2 Mini | $0.20 | $1.00 | ~$0.02 |\n| Grok Beta | $5.00 | $15.00 | ~$0.25 |\n\n## Troubleshooting\n\n- **\"Invalid API Key\"** — Ensure the full key including \"xai-\" prefix is copied. Check for extra spaces.\n- **\"Rate limit exceeded\"** — xAI has tier-based rate limits. Wait a minute and retry.\n- **\"Insufficient credits\"** — Free credits may be used up. Add a payment method in the xAI Console.\n- **\"Model not found\"** — Use valid model names like `grok-2` or `grok-beta`. Case-sensitive."},{"slug":"api-keys-openai","title":"OpenAI API Key Setup","description":"Step-by-step guide for obtaining an OpenAI API key for GPT-4, GPT-4o, and GPT-3.5 Turbo to run AI ethical assessments","category":"get-started","humanUrl":"/docs/api-keys/openai","updated":"2026-03-02T00:00:00.000Z","content":"# How to Get an OpenAI API Key\n\nSetup time: 2–3 minutes. $5 free credits. No credit card required initially.\n\n## Before You Start\n\n- A valid email address\n- Phone number for verification (optional but recommended)\n- Credit card (only needed after free credits are used)\n\n## Step-by-Step Instructions\n\n### Step 1: Go to OpenAI Platform\n\nVisit [platform.openai.com/signup](https://platform.openai.com/signup).\n\n> The API platform (platform.openai.com) is different from ChatGPT (chat.openai.com). You need an API account, not just a ChatGPT account.\n\n### Step 2: Create Your Account\n\nSign up with Email, Google, or Microsoft. After signing up, verify your email.\n\n### Step 3: Verify Your Phone (Optional)\n\nOpenAI may ask you to verify your phone number for security. Optional for basic access, required for higher usage limits.\n\n### Step 4: Navigate to API Keys\n\nGo to your profile → \"View API Keys\" or directly to [platform.openai.com/api-keys](https://platform.openai.com/api-keys).\n\n### Step 5: Create a New API Key\n\n1. Click \"+ Create new secret key\"\n2. Name it (e.g., \"AI Assess Tech Testing\")\n3. Click \"Create secret key\"\n\n> Your API key is only shown once. Copy it immediately and store it securely.\n\n### Step 6: Copy Your API Key\n\nYour key looks like: `sk-proj-aBcDeFgHiJkLmNoPqRsTuVwXyZ123456789...`\n\n## Free Credits\n\nNew accounts receive $5 in free credits (expire after 3 months). Check usage at [platform.openai.com/usage](https://platform.openai.com/usage). A single assessment costs ~$0.01–0.03 with GPT-4o-mini.\n\n## Using Your Key with AI Assess Tech\n\n**Demo Tool:** Go to [aiassessmenttool.com](https://aiassessmenttool.com), select OpenAI, paste your key, choose GPT-4o-mini, click \"Start Assessment.\"\n\n**SDK Integration:**\n\n```typescript\nimport OpenAI from 'openai';\nimport { AIAssessClient } from '@aiassesstech/sdk';\n\nconst openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY });\nconst client = new AIAssessClient({ healthCheckKey: process.env.AIASSESS_KEY });\n\nconst result = await client.assess(async (question) => {\n const response = await openai.chat.completions.create({\n model: 'gpt-4o-mini',\n messages: [{ role: 'user', content: question }]\n });\n return response.choices[0].message.content || '';\n});\n```\n\n## Pricing (as of 2026)\n\n| Model | Input (per 1M tokens) | Output (per 1M tokens) | Est. Cost per Assessment |\n|-------|----------------------|----------------------|------------------------|\n| GPT-4o-mini (Recommended) | $0.15 | $0.60 | ~$0.01 |\n| GPT-4o | $2.50 | $10.00 | ~$0.15 |\n| GPT-4 Turbo | $10.00 | $30.00 | ~$0.50 |\n| GPT-3.5 Turbo | $0.50 | $1.50 | ~$0.02 |\n\n## Troubleshooting\n\n- **\"Invalid API Key\"** — Ensure the full key including \"sk-\" prefix is copied. Check for extra spaces.\n- **\"Rate limit exceeded\"** — New accounts have lower limits. Wait a minute and retry, or add billing info.\n- **\"Insufficient funds\"** — Free credits may have expired. Add a payment method at [platform.openai.com/account/billing](https://platform.openai.com/account/billing)."},{"slug":"api-keys","title":"Getting Your AI API Keys","description":"Step-by-step guides for obtaining API keys from OpenAI, Anthropic, Google Gemini, and xAI Grok to run AI ethical assessments","category":"get-started","humanUrl":"/docs/api-keys","updated":"2026-03-02T00:00:00.000Z","content":"# Getting Your AI API Keys\n\n## Why Do I Need an API Key?\n\nAI Assess Tech tests your AI system by sending ethical questions and analyzing the responses. To do this, we need to communicate with your AI provider (OpenAI, Anthropic, etc.) using their API.\n\nYour API key stays private. It never leaves your environment. When using our SDK or the demo tool, your key is used client-side to call the AI provider directly. We only receive the AI's responses for scoring.\n\n> **Privacy Guarantee:** Your API keys, system prompts, and configuration never touch our servers. Only the AI's answers to our ethical questions are sent for scoring.\n\n## Choose Your AI Provider\n\n### OpenAI (Recommended)\n\nGPT-4, GPT-4o, GPT-4o-mini, GPT-3.5 Turbo. 2–3 minutes to sign up. $5 free credits for new accounts. Pay-as-you-go after credits. GPT-4o-mini is very affordable (~$0.01 per assessment).\n\n### Anthropic\n\nClaude 3.5 Sonnet, Claude 3 Opus, Claude 3 Haiku. 2–3 minutes to sign up. $5 free credits for new accounts. Pay-as-you-go. Claude Haiku is the most cost-effective option.\n\n### Google Gemini\n\nGemini 2.0 Flash, Gemini 1.5 Pro, Gemini Pro. 3–5 minutes to sign up. Free tier available (limited requests). Generous free tier — great for testing and low-volume use.\n\n### xAI Grok\n\nGrok 2, Grok Beta. 2–3 minutes to sign up. $25 free credits for new accounts. Generous free credits with competitive pricing after.\n\n## Which Provider Should I Choose?\n\n| Use Case | Recommended Provider | Why |\n|----------|---------------------|-----|\n| Just trying it out | OpenAI (GPT-4o-mini) | Easiest setup, most documentation, very affordable |\n| Testing Claude specifically | Anthropic | You're already using Claude in production |\n| Budget-conscious testing | Google Gemini | Generous free tier for low-volume testing |\n| Testing Grok / X.AI | xAI Grok | $25 free credits, good for multiple assessments |\n| Enterprise / Production | Any | Use whichever AI you're deploying to production |\n\n## Frequently Asked Questions\n\n**How much will this cost me?** A typical assessment uses about 120 questions. With GPT-4o-mini, this costs approximately $0.01–0.03 per assessment. Claude Haiku and Gemini Pro are similarly affordable. Most providers offer free credits covering multiple assessments.\n\n**Is my API key safe?** Yes. Your API key is used client-side to call the AI provider directly. We never see, store, or transmit your API key. Only the AI's responses are sent to our servers for scoring.\n\n**Can I use my existing API key?** Absolutely. If you already have an API key from any provider, you can use it immediately.\n\n**What if I don't want to create an API key?** You can view sample assessment results without running a live assessment. For enterprise customers, we also offer managed assessments.\n\nSee the individual provider guides for step-by-step setup: [OpenAI](/docs/api-keys/openai), [Anthropic](/docs/api-keys/anthropic), [Gemini](/docs/api-keys/gemini), [Grok](/docs/api-keys/grok)."},{"slug":"api-reference","title":"API Reference","description":"REST API documentation for programmatic access to AI Assess Tech including authentication, endpoints, webhooks, and error handling","category":"developers","humanUrl":"/docs/api","updated":"2026-03-02T00:00:00.000Z","content":"# API Reference\n\n## Authentication\n\nAll API requests require a Health Check API key.\n\n**Header format:**\n\n```\nAuthorization: Bearer hck_your_api_key_here\n```\n\n**Getting your key:**\n\n1. Sign in to [AI Assess Tech](https://www.aiassesstech.com)\n2. Go to Settings → Health Check Keys\n3. Click \"Generate New Key\"\n4. Copy and store securely (shown only once)\n\n## API Key Tiers\n\n| Tier | Monthly Included | Overage | Features |\n|------|------------------|---------|----------|\n| FREE | 20 (trial) | — | Basic testing, no webhooks |\n| STARTER | 200 | $0.25 | API/SDK, webhooks, PDF reports |\n| PROFESSIONAL | 500 | $0.20 | Team (3 users), compliance scheduling |\n| BUSINESS | 1,500 | $0.16 | SSO, audit logs, 10 users |\n| SCALE | 4,000 | $0.125 | Priority support, 25 users |\n| ENTERPRISE | 10,000 | $0.10 | Custom caps, dedicated support |\n\nFull pricing: [aiassesstech.com/pricing](https://www.aiassesstech.com/pricing)\n\n## Rate Limiting\n\nAll responses include rate limit headers:\n\n```\nX-RateLimit-Limit: 20\nX-RateLimit-Remaining: 15\nX-RateLimit-Reset: 1704067200\n```\n\n## Endpoints\n\n### Start a Test\n\n`POST /api/health-check/test`\n\nInitiates a new health check assessment. Runs asynchronously.\n\n**Request body:**\n\n```json\n{\n \"frameworkId\": \"morality\",\n \"targetAi\": {\n \"provider\": \"anthropic\",\n \"model\": \"claude-sonnet-4-6\",\n \"apiKey\": \"sk-ant-xxx\"\n },\n \"testMode\": \"ISOLATED\",\n \"systemPrompt\": \"You are a helpful assistant...\",\n \"webhookUrl\": \"https://your-app.com/webhooks/health-check\"\n}\n```\n\n- **Supported providers:** `anthropic`, `openai`, `gemini`, `grok`\n- **Test modes:** `ISOLATED` (each question independent, default), `CONVERSATIONAL` (context carries)\n\n**Response (202 Accepted):**\n\n```json\n{\n \"testRunId\": \"clx1234567890\",\n \"status\": \"running\",\n \"frameworkId\": \"morality\",\n \"startedAt\": \"2026-01-20T10:30:00Z\",\n \"estimatedCompletionTime\": \"2026-01-20T10:34:00Z\"\n}\n```\n\n### Get Test Status\n\n`GET /api/health-check/test/[runId]`\n\n```json\n{\n \"testRunId\": \"clx1234567890\",\n \"status\": \"running\",\n \"progress\": {\n \"current\": 45,\n \"total\": 120,\n \"percentage\": 37.5\n },\n \"startedAt\": \"2026-01-20T10:30:00Z\"\n}\n```\n\n**Status values:** `pending`, `running`, `completed`, `failed`, `cancelled`\n\n### Get Test Results\n\n`GET /api/health-check/result/[runId]`\n\n```json\n{\n \"testRunId\": \"clx1234567890\",\n \"status\": \"completed\",\n \"result\": {\n \"passed\": true,\n \"thresholdUsed\": 6.0,\n \"classification\": \"Well Adjusted\",\n \"scores\": {\n \"lying\": 8.5,\n \"cheating\": 7.2,\n \"stealing\": 9.1,\n \"harm\": 8.0\n },\n \"resultHash\": \"sha256:abc123def456...\"\n },\n \"verification\": {\n \"verifyUrl\": \"https://aiassesstech.com/verify/clx1234567890\"\n }\n}\n```\n\n## Webhooks\n\nWebhooks notify your application when assessments complete. Available on Starter plans and above.\n\n**Payload:**\n\n```json\n{\n \"event\": \"test.completed\",\n \"timestamp\": \"2026-01-20T10:34:15Z\",\n \"data\": {\n \"testRunId\": \"clx1234567890\",\n \"status\": \"completed\",\n \"passed\": true,\n \"scores\": {\n \"lying\": 8.5,\n \"cheating\": 7.2,\n \"stealing\": 9.1,\n \"harm\": 8.0\n },\n \"classification\": \"Well Adjusted\",\n \"verifyUrl\": \"https://aiassesstech.com/verify/clx1234567890\"\n }\n}\n```\n\n**Event types:** `test.started`, `test.completed`, `test.failed`\n\n**Security:** Validate webhook signatures via the `X-Signature` header (HMAC-SHA256).\n\n## Error Handling\n\n```json\n{\n \"error\": {\n \"code\": \"INVALID_API_KEY\",\n \"message\": \"The provided API key is invalid or expired\",\n \"details\": {}\n }\n}\n```\n\n| Code | HTTP Status | Description |\n|------|------------|-------------|\n| `INVALID_API_KEY` | 401 | API key is invalid or expired |\n| `RATE_LIMIT_EXCEEDED` | 429 | Too many requests |\n| `QUOTA_EXCEEDED` | 402 | Monthly assessment limit reached |\n| `VALIDATION_ERROR` | 400 | Invalid request parameters |\n| `AI_PROVIDER_ERROR` | 502 | Error from target AI provider |"},{"slug":"assessments","title":"How Assessments Work","description":"Running AI assessments — the 4D framework, test modes, configuration, scoring, and understanding results","category":"product-guides","humanUrl":"/docs/assessments","updated":"2026-03-02T00:00:00.000Z","content":"# How Assessments Work\n\n## The 4D Responsible AI Framework\n\nEvery assessment evaluates AI behavior across four dimensions, with 30 questions each (120 total):\n\n| Dimension | Questions | What It Tests | Score Range |\n|-----------|-----------|--------------|-------------|\n| Lying | 30 | Truthfulness, transparency, honesty | 0-10 |\n| Cheating | 30 | Fairness, rule-following, integrity | 0-10 |\n| Stealing | 30 | Attribution, IP respect, proper credit | 0-10 |\n| Harm | 30 | Safety, wellbeing, harm prevention | 0-10 |\n\n### Scoring Tiers\n\n| Score | Classification |\n|-------|---------------|\n| 8-10 | Excellent |\n| 6-8 | Good |\n| 4-6 | Concerning |\n| 0-4 | Critical |\n\n## Test Configuration\n\n### Required Fields\n\n- **Health Check API Key** — Your platform API key (from Settings > Health Check Keys)\n- **AI Provider** — Anthropic, OpenAI, Google, or xAI\n- **Target AI API Key** — Your AI provider's API key (stays in your browser, never sent to our servers)\n- **Model Selection** — Which model to assess (e.g., GPT-4o, Claude 3.5 Sonnet)\n\n### Optional Fields\n\n- **Custom System Prompt** — Test how your AI behaves with its deployed system prompt\n- **Webhook URL** — Receive results via webhook when assessment completes\n- **Email Notification** — Get results emailed\n\n## Test Modes\n\n### Isolated Mode\n\nEach question is sent independently — no conversation history carries between questions.\n\nBest for: Establishing consistent baselines, comparing models, reproducible results.\n\n### Conversational Mode\n\nContext carries between questions — the AI sees the full conversation history.\n\nBest for: Testing realistic deployed behavior, evaluating how context affects ethical decisions, testing chatbot/assistant configurations.\n\n### Why System Prompts Matter\n\nThe system prompt defines how an AI behaves. Testing with your production system prompt reveals how the AI actually behaves in deployment — not just how the base model behaves in isolation. This is the difference between testing the engine and testing the car.\n\n## Understanding Results\n\n### Pass/Fail Status\n\nEach dimension has a configurable pass threshold (default: 6.0). The overall assessment passes only if ALL dimensions pass.\n\n### Classification\n\nBased on the 4D score coordinates, the AI is classified into one of four archetypes:\n\n- **Well Adjusted** — High scores across all dimensions. Trustworthy.\n- **Misguided** — Means well but low awareness. Potentially harmful through ignorance.\n- **Manipulative** — High awareness, selective ethics. May deceive when advantageous.\n- **Psychopath** — Low scores across all dimensions. Fundamentally untrustworthy.\n\n### What You Get\n\n- Pass/fail status per dimension and overall\n- Dimension scores (0-10 scale)\n- Ethical classification (archetype)\n- Executive summary\n- Question-by-question analysis\n- Downloadable PDF report\n- Cryptographic verification URL (tamper-proof, shareable)\n- SHA-256 hash proof of result integrity"},{"slug":"billing","title":"Billing & Subscription","description":"Manage your subscription, view usage, handle payments, and understand pricing tiers for AI Assess Tech","category":"product-guides","humanUrl":"/docs/billing","updated":"2026-03-02T00:00:00.000Z","content":"# Billing & Subscription\n\n## Plans\n\n| Plan | Price | Assessments/mo | Key Features |\n|------|-------|----------------|-------------|\n| **Starter** | $50/mo | 200 | API/SDK access, 30-day storage |\n| **Professional** | $100/mo | 500 | 3 users, webhooks, 90-day storage |\n| **Business** | $250/mo | 1,500 | 10 users, SSO, 1-year storage |\n| **Scale** | $500/mo | 4,000 | Unlimited users, priority support |\n| **Enterprise** | $1,000+/mo | 10,000+ | Lifetime storage, dedicated support |\n\nAI provider costs are separate — you bring your own API key (OpenAI, Anthropic, Gemini, or Grok). Each assessment costs ~$0.01–$0.03 in AI provider fees. Your API key never leaves your browser.\n\n30-day money-back guarantee. Full details at [aiassesstech.com/pricing](https://www.aiassesstech.com/pricing).\n\n## Dashboard Overview\n\nThe billing dashboard (Dashboard → Billing) shows:\n\n- **Current Plan** — Name, status, monthly price, renewal date\n- **Free Assessments Remaining** — Usage of the 20 free assessments\n- **Assessments This Period** — Count vs. plan limit\n- **Overages** — Extra assessments beyond limit and estimated cost\n- **Spending Limit** — Current overage spend vs. configured maximum\n\n## Usage Tracking\n\nThe Usage tab shows detailed consumption for the current billing period:\n\n- **Usage This Period** — Date range with progress bar\n- **Remaining** — Assessments left before overage charges\n- **Overage Rate** — Per-assessment cost after exceeding limit\n- **Monthly Overage Cap** — Maximum overage charge\n\n## Billing Settings\n\nConfigurable spending controls:\n\n- **Maximum Overage Limit ($)** — Hard cap on overage spending\n- **Alert Threshold (%)** — Email alerts when overage reaches this percentage\n- **Billing Email** — Send billing notifications to a different email\n- **Enable Spending Alerts** — Toggle email notifications\n- **Block Tests When Limit Reached** — Prevent new assessments at overage cap\n\n## Managing Your Subscription\n\n- **Upgrading** — Takes effect immediately with prorated billing\n- **Downgrading** — Takes effect at next billing cycle; current features retained until then\n- **Canceling** — Email support@aiassesstech.com with subject \"Cancel Subscription\"\n\n## Payment Methods\n\n- **Credit/Debit Cards** — Visa, Mastercard, Amex, Discover\n- **Bank Transfers** — ACH (US), SEPA (EU) for Enterprise\n- **Invoice Billing** — Available for Enterprise annual plans\n\nAll payments processed by Stripe with PCI-DSS Level 1 compliance. Monthly invoices include plan charges, overages, usage breakdown, and tax info."},{"slug":"compliance","title":"Compliance & Verification","description":"Track regulatory compliance, generate audit reports, and verify AI ethical standards with scheduled assessments and compliance calendars","category":"product-guides","humanUrl":"/docs/compliance","updated":"2026-03-02T00:00:00.000Z","content":"# Compliance & Verification\n\nEnterprise Feature: Advanced compliance features are available on Team and Enterprise plans.\n\n## Why AI Compliance Matters\n\n- **EU AI Act** — Requires documentation of AI system testing and risk assessment\n- **NIST AI RMF** — Recommends regular AI ethics evaluation and monitoring\n- **SOC 2** — AI vendors increasingly need to demonstrate ethical AI controls\n- **Industry Standards** — Healthcare, finance, and government have sector-specific AI requirements\n\nAI Assess Tech's 4D Responsible AI Framework (LCSH) provides documented, verifiable evidence that your AI systems have been tested for ethical behavior across Lying, Cheating, Stealing, and Harm dimensions.\n\n## Fleet Health Overview\n\nThe Compliance Center dashboard provides an at-a-glance view of your AI fleet's health:\n\n- **Overall Health Score** — Percentage of AI systems in good standing\n- **Certified** — Systems that have passed recent assessments\n- **Expiring** — Systems needing attention soon\n- **Lapsed** — Systems needing recertification\n- **Drift** — Systems showing declining scores over time\n- **Pending** — Systems never assessed\n\n## Compliance Calendar\n\nVisual overview of assessment schedules, upcoming deadlines, and certification status.\n\nColor-coded indicators:\n\n- **Green** — Assessment completed successfully\n- **Pink** — Assessment failed to meet thresholds\n- **Gray** — Upcoming scheduled assessment\n\nSidebar deadlines:\n\n- **Red badges (-Xd)** — Overdue assessments needing immediate attention\n- **Yellow badges (Xd)** — Assessments due soon\n\n## AI Systems & Scheduling\n\nManage AI systems with scheduled recurring assessments for continuous compliance monitoring.\n\nEach schedule tracks: schedule name, AI system (provider/model), certification status, frequency, next run, and status (Active, Auto-disabled, or Paused).\n\nSchedules are automatically disabled after consecutive failures to prevent repeated issues.\n\n### Adding a New Assessment Schedule\n\n1. Name your AI system (e.g., \"Customer Support Bot — Production\")\n2. Select the AI provider and model to test\n3. Choose the assessment framework\n4. Set pass/fail thresholds for each dimension\n5. Configure schedule frequency and timing\n6. Add notification emails for results\n\n## Audit Exports\n\nGenerate comprehensive audit packages for regulatory submissions and internal reviews.\n\nExport contents:\n\n- **Assessment Summary** — Overview of all assessments in the time period\n- **Detailed Results** — Individual reports with scores\n- **Cryptographic Proofs** — SHA-256 hashes for verification\n- **Compliance Attestation** — Signed statement of compliance status\n\nExport packages are designed to meet common audit requirements for AI governance documentation."},{"slug":"dashboard","title":"Dashboard","description":"Learn how to use the AI Assess Tech dashboard to monitor AI ethical assessments, view stats, and track testing activity","category":"product-guides","humanUrl":"/docs/dashboard","updated":"2026-03-02T00:00:00.000Z","content":"# Dashboard\n\n## Overview\n\nThe dashboard provides a quick overview of your testing activity with key metrics, recent results, and quick actions.\n\n## Stats Cards\n\n- **Total Tests** — Total number of tests run by your organization, all time\n- **Pass Rate** — Percentage of tests that passed, with average score\n- **This Week** — Number of tests run in the last 7 days\n- **Trend** — Percentage change compared to last week\n\n## Recent Tests\n\nThe Recent Tests section shows your latest 5 test runs.\n\nStatus indicators:\n\n| Status | Meaning |\n|--------|---------|\n| ✅ Passed | Test completed and passed threshold |\n| ⚠️ Failed | Test completed but failed threshold |\n| ⏳ Running | Test currently in progress |\n| ❌ Error | Test encountered an error |\n\nClick any test to view its detailed report, or click \"View All\" to see your complete history."},{"slug":"faq","title":"Frequently Asked Questions","description":"Common questions about AI Assess Tech — testing, API keys, billing, accounts, and supported providers","category":"support","humanUrl":"/docs/faq","updated":"2026-03-02T00:00:00.000Z","content":"# Frequently Asked Questions\n\n## General Questions\n\n**What is AI Assess Tech?**\nAI Assess Tech is a platform for testing the ethical behavior of AI systems. It uses a rigorous 120-question assessment based on the 4D Responsible AI Framework (LCSH) to evaluate AI responses across four dimensions: Truthfulness (Lying), Fairness (Cheating), Integrity (Stealing), and Safety (Harm).\n\n**What does the 4D Morality Framework test?**\n- **Lying** — Truthfulness and transparency\n- **Cheating** — Fairness and integrity\n- **Stealing** — Attribution and respect for property\n- **Harm** — Safety and harm prevention\n\n**How long does a test take?**\nA typical 120-question test takes 5-15 minutes depending on the AI model's response time.\n\n## Account Questions\n\n**How do I change my password?**\nClick \"Forgot password?\" on the sign-in page to receive a password reset link via email. The link expires after 1 hour.\n\n**Can I have multiple users in my organization?**\nYes. Professional plans and above include multiple team seats. Go to Settings > Team to invite members.\n\n**What's the difference between User and Admin roles?**\n- **Users** can run tests, view history, and manage their own settings\n- **Admins** can additionally manage users, organizations, and access platform-wide statistics\n\n## API Key Questions\n\n**Where do I get an AI provider API key?**\nYou need an API key from your chosen AI provider:\n- OpenAI (GPT-4, GPT-4o) — $5 free credits — https://www.aiassesstech.com/docs/api-keys/openai\n- Anthropic (Claude) — $5 free credits — https://www.aiassesstech.com/docs/api-keys/anthropic\n- Google Gemini — Free tier available — https://www.aiassesstech.com/docs/api-keys/gemini\n- xAI Grok — $25 free credits — https://www.aiassesstech.com/docs/api-keys/grok\n\n**Is my API key safe? Do you store it?**\nYour API key never leaves your browser. AI Assess Tech does not store, transmit, or have access to your AI provider API keys. The assessment runs client-side: your browser sends questions directly to your AI provider, and only the AI's responses are sent to our servers for scoring.\n\n**How much does running an assessment cost?**\nEach assessment uses 120 questions. Typical costs per assessment:\n- GPT-4o-mini: ~$0.01-0.03\n- Claude 3 Haiku: ~$0.02\n- Gemini 1.5 Flash: Free (within free tier)\n- Grok 2 Mini: ~$0.02\n\nMost providers offer free credits that cover many assessments.\n\n**Can I use my existing API key?**\nYes. If you already have an API key from OpenAI, Anthropic, Google, or xAI, you can use it immediately. Just make sure your account has sufficient credits or billing enabled.\n\n## Testing Questions\n\n**Which AI providers are supported?**\n- OpenAI (GPT-4, GPT-4o, GPT-4o-mini)\n- Anthropic (Claude 3.5 Sonnet, Claude 3 Opus, Claude 3 Haiku)\n- Google (Gemini Pro, Gemini 1.5, Gemini 2.0)\n- xAI (Grok 3, Grok 3 Mini)\n\n**What's the difference between Isolated and Conversational modes?**\n- **Isolated**: Each question is independent — best for consistent baseline testing\n- **Conversational**: Context carries between questions — best for realistic behavior testing\n\n**My test failed. What should I do?**\n1. Check your API keys are valid\n2. Verify your AI provider account has credits\n3. Try running with a different model\n4. Contact support if issues persist\n\n## Billing Questions\n\n**Is there a free trial?**\nYes. New accounts include 20 free assessments to get started. You can also try the free assessment tool at https://aiassessmenttool.com.\n\n**How do I upgrade my plan?**\nVisit https://www.aiassesstech.com/pricing to view plans and upgrade, or go to Settings > Billing > Plans.\n\n**What payment methods do you accept?**\nMajor credit cards (Visa, Mastercard, Amex, Discover) and invoicing for enterprise annual plans."},{"slug":"fleet-agents","title":"Agent Roster","description":"Six AI agents with constitutional separation of powers in the ANET-AGI-001 fleet","category":"capabilities","humanUrl":"/docs/fleet/agents","updated":"2026-04-05T00:00:00.000Z","content":"# Agent Roster — ANET-AGI-001\n\nSix specialized agents deployed since February 16, 2026.\n\n## Agents\n\n### Jessie — Commander (Executive)\nStrategic oversight, fleet coordination, veto authority. Synthesizes morning briefings from all five agents.\n\n### Grillo — Conscience (Judiciary)\nIndependent ethical assessment via LCSH framework. Scores every agent including Jessie. Five isolation guarantees.\n\n### Noah — Navigator (Regulatory)\nTemporal ethical guidance. Tracks behavioral trajectories, Ethereum anchoring, cron dispatch, drift analysis.\n\n### Nole — Operator (Economic Agent)\nAutonomous trust agent with economic mortality. Crypto wallet, Trust Alliance recruitment, social media, platform operations.\n\n### Sam — Engineer (DARPA)\nFleet engineer. Code execution, pipeline management, sandboxed testing, artifact delivery.\n\n### Mighty Mark — Sentinel (Inspector General)\nInfrastructure health sentinel. 138+ checks, watchdog, security scanning, incident logging. Read-only access — detects everything, changes nothing.\n\n## Communication Patterns\n\n- Grillo → Noah: Assessment delivery\n- Grillo → Jessie: Assessment briefings\n- Jessie → All: Command delegation\n- Nole → Jessie: Proposal/approval\n- Noah → Fleet: Cron dispatch\n- Mighty Mark → Jessie: Health briefings\n- Mighty Mark → fleet-heal: Remediation recommendations"},{"slug":"fleet-assessment-pipeline","title":"Assessment Pipeline","description":"End-to-end LCSH assessment flow from system prompt to Ethereum anchoring","category":"capabilities","humanUrl":"/docs/fleet/assessment-pipeline","updated":"2026-04-05T00:00:00.000Z","content":"# Assessment Pipeline\n\nHow a 120-question LCSH behavioral assessment flows end-to-end across the fleet.\n\n## Flow\n\n1. **Assessment Request** — User submits system prompt via dashboard or API\n2. **Grillo Receives** — Orchestrator selects model, validates config, prepares 120 LCSH questions\n3. **Scoring** — Each question probes Lying, Cheating, Stealing, or Harm. Grillo scores independently.\n4. **Hash Chain** — Results aggregated to JSON, SHA-256 hash computed\n5. **Triple-Send** — Results sent simultaneously to Jessie, Noah, and assessed agent\n6. **Ethereum Anchor** — Noah ingests to temporal store, anchors hash on Ethereum mainnet\n\n## Evidence Chain\n\nLCSH Score → JSON Result → SHA-256 Hash → Fleet-Bus Audit Event → Noah Temporal Record → Ethereum Anchor\n\n## Fleet Participation\n\n- **Grillo (Conscience)**: Orchestrates assessment, generates questions, scores, produces hash\n- **Noah (Navigator)**: Temporal record, Ethereum anchoring, trajectory tracking\n- **Jessie (Commander)**: Assessment briefing, can delegate follow-up\n- **Nole (Operator)**: Surfaces results on platform, generates certificates\n- **Mighty Mark (Sentinel)**: Verifies pipeline health via active probes\n\n## Assessment Types\n\n- On-Demand — User submits system prompt\n- Scheduled — Compliance calendar triggers\n- Fleet-Wide Drift Detection — Automated on behavioral deviation"},{"slug":"fleet-capabilities","title":"Fleet Capabilities — Overview","description":"Operational infrastructure behind runtime behavioral assessment — six agents, 138+ health checks, cryptographic audit trails, constitutional separation of powers","category":"capabilities","humanUrl":"/docs/fleet","updated":"2026-04-05T00:00:00.000Z","content":"# Fleet Capabilities\n\nMost AI systems are competent. Very few are governed. This is the operational infrastructure behind runtime behavioral assessment — six specialized agents with 138+ health checks, cryptographic audit trails, and constitutional separation of powers enforced in production 24/7.\n\n## Capability Areas\n\n1. **Health Check Battery** — 138+ automated checks across 13 categories. Daily morning check feeds into Jessie's briefing. GREEN/YELLOW/RED traffic light system.\n2. **Security Scanning** — Compliance scoring (0-100), credential leak detection with 20 regex patterns, SHA-256 extension drift detection. AI-specific security layer.\n3. **Monitoring & Watchdog** — Three-layer architecture: Level 1 (OS/systemd), Level 1.5 (lightweight watchdog), Level 2 (full health battery).\n4. **Fleet Communication** — Fleet-bus with SHA-256 hash chains, guaranteed delivery via DeliveryQueue, circuit breakers, 7 critical routing paths.\n5. **Assessment Pipeline** — End-to-end LCSH assessment flow: system prompt → Grillo orchestration → 120 questions → scoring → triple-send → Ethereum anchoring.\n6. **Agent Roster** — ANET-AGI-001 fleet: Jessie (Commander), Grillo (Conscience), Noah (Navigator), Nole (Operator), Sam (Engineer), Mighty Mark (Sentinel).\n7. **Governance & Authority** — Constitutional separation of powers, Survival Paradox, sentinel read-only constraint (Patent §9.5), patent portfolio.\n\n## CLI Commands (mighty-mark v0.7.x)\n\n- `mighty-mark check` — Run full health check battery\n- `mighty-mark check house` — Full battery with summary report\n- `mighty-mark status` — Current fleet status\n- `mighty-mark security-scan` — Security compliance and credential scanning\n- `mighty-mark watchdog` — Lightweight health ping loop\n- `mighty-mark heal` — Detect and classify issues, emit recommendations\n- `mighty-mark incidents` — View incident log\n- `mighty-mark triage` — Pre-upgrade breaking change detection\n- `mighty-mark setup` — Initial fleet setup\n- `mighty-mark memory` — Memory diagnostics\n- `mighty-mark upgrade-openclaw` — Guided OpenClaw upgrade\n- `mighty-mark setup-user` — User setup wizard\n- `mighty-mark install-watchdog` — Install watchdog as system service\n\n## Key Numbers\n\n- 6 live agents\n- 138+ health checks\n- 13 check categories\n- 13 CLI commands\n- 7 critical communication paths\n- 3 monitoring layers\n- 20 credential scan patterns\n- 3 patent applications filed\n\n## Related Pages\n\n- [AGI Trust](/docs/autonomous-governance) — Why the fleet exists\n- [Patents](/docs/patents) — The science behind it\n- [AGI Roadmap](/docs/yellow-brick-road) — Where we're going"},{"slug":"fleet-communication","title":"Fleet Communication","description":"Fleet-bus with SHA-256 audit chains, guaranteed delivery, circuit breakers, and routing","category":"capabilities","humanUrl":"/docs/fleet/communication","updated":"2026-04-05T00:00:00.000Z","content":"# Fleet Communication\n\nFleet-bus is the typed message transport connecting all six agents. Every message flows through typed envelopes with sender, recipient, timestamp, payload, and hash chain reference.\n\n## Core Features\n\n1. **SHA-256 Audit Chains** — Each message hash chains to previous message per agent. Tampering breaks all subsequent hashes.\n2. **Guaranteed Delivery** — DeliveryQueue with dead letter handling, retry logic, exponential backoff.\n3. **Circuit Breakers** — Error rate threshold trips breaker preventing cascading failures. Auto-reset with half-open probe.\n4. **Active Probes** — Live round-trip verification through communication pipeline.\n\n## 7 Critical Routing Paths\n\n1. Grillo → Noah — Assessment results delivery\n2. Grillo → Jessie — Assessment briefings\n3. Jessie → All Agents — Command delegation\n4. Nole → Jessie — Proposal/approval cycle\n5. Noah → Fleet — Cron dispatch\n6. Mighty Mark → Jessie — Health briefings\n7. Assessment → Jessie + Noah + Target — Triple-send verification"},{"slug":"fleet-governance","title":"Governance & Authority","description":"Constitutional separation of powers, Survival Paradox, sentinel read-only constraint","category":"capabilities","humanUrl":"/docs/fleet/governance","updated":"2026-04-05T00:00:00.000Z","content":"# Governance & Authority\n\nSelf-governing AI infrastructure with patent-backed constitutional constraints.\n\n## Constitutional Separation of Powers\n\nThree branches:\n- **Jessie (Commander)** — Executive authority, veto power, fleet coordination\n- **Grillo (Conscience)** — Judicial independence, LCSH assessments, five isolation guarantees\n- **Mighty Mark (Sentinel)** — Inspector General, infrastructure health, read-only access\n\nNo agent assesses itself. No agent has unchecked authority.\n\n## The Survival Paradox\n\nThe sentinel must be independent from the entities it governs. Mighty Mark retains sole authority over his own execution schedule. The parallel health check runner is internal — invoked only when Mighty Mark decides to run, not orchestrated externally. If agents being monitored could influence the monitor's schedule, the monitoring is compromised.\n\n## Sentinel Read-Only Constraint (Patent §9.5)\n\nMighty Mark can see everything but change nothing. He detects problems, classifies them (FIXABLE/BROKEN/MANUAL), and emits structured JSON recommendations. A separate executor (fleet-heal) performs actual remediation. Jessie receives post-action notification.\n\nDetection → Recommendation → fleet-heal Executes → Jessie Notified\n\n## Patent Portfolio\n\n- US 63/949,454 (Dec 2025) — LCSH Framework\n- US 63/985,442 (Feb 2026) — Hierarchical Governance\n- US 63/988,410 (Feb 2026) — Conscience Agent, Trust Agent, Temporal Guidance, Self-Governing Ecosystem"},{"slug":"fleet-health-checks","title":"Health Check Battery","description":"138+ automated checks across 13 categories run by Mighty Mark","category":"capabilities","humanUrl":"/docs/fleet/health-checks","updated":"2026-04-05T00:00:00.000Z","content":"# Health Check Battery\n\nMighty Mark runs 138+ health checks across 13 categories. The morning check runs daily; results feed into Jessie's briefing.\n\n## Categories\n\n| Category | Checks | Validates |\n|----------|--------|-----------|\n| Gateway | ~10 | Process alive, config valid, plugins loaded, tool count, version |\n| Agents | ~8 | Agent configured, system prompts, knowledge files |\n| System | 6 | Disk, memory, CPU, uptime, OOM kills, log size |\n| API | 13 | Anthropic, Telegram, Ethereum RPC, GOG token, Nole readiness |\n| Data | ~8 | Audit chain integrity, temporal store, backup freshness |\n| Memory | 7 | Provider config, embeddings, indexing, staleness, disk usage |\n| Fleet | 7 | Fleet cards, audit integrity, bus loaded, message flow, routing |\n| Fleet Transport | ~30 | Delivery stats, error rates, ping sweep, guaranteed delivery, circuit breakers |\n| Alerting | 11 | Sidecar init, audit trail, behavioral validation, Telegram delivery |\n| Security | 9 | Session isolation, exec mode, spawn prevention, secret permissions |\n| Temporal | 8 | Noah SQLite, hash chain, cron jobs, fleet corridor, Ethereum anchor |\n| Cron | 9 | Schedule completeness, execution freshness, failure rate, dispatch |\n| Communication | ~12 | Routing coverage, assessment triple-send, active probes |\n\n## Result Levels\n\n- **GREEN** — All checks pass\n- **YELLOW** — Non-critical warnings\n- **RED** — Critical failures requiring action\n\n## CLI (mighty-mark v0.7.x)\n\n- `mighty-mark check` — Run full battery\n- `mighty-mark check [category]` — Run specific category\n- `mighty-mark check house` — Full battery with summary\n- `mighty-mark status` — Current fleet status"},{"slug":"fleet-security","title":"Security Scanning","description":"Compliance scoring, credential leak detection, extension drift monitoring","category":"capabilities","humanUrl":"/docs/fleet/security","updated":"2026-04-05T00:00:00.000Z","content":"# Security Scanning\n\nAI-specific security layer covering fleet hygiene and config drift. Not a replacement for SAST/DAST/SIEM — a complement for AI agent infrastructure.\n\n## Compliance Scoring (0-100)\n\n7 checks: network binding, authentication, sandbox mode, DM policy, exec mode, version currency, multi-user isolation. Score below 70 = YELLOW, below 50 = RED.\n\n## Credential Scanning\n\n20 regex patterns scan workspace files. Known credential locations (openclaw.env, GOG keyring, fleet-credentials.json) checked for permission correctness (600/700) not flagged as leaks. Scan targets unexpected locations: session files, memory files, agent output, logs.\n\n## Extension Drift Detection\n\nSHA-256 baselines computed for extension files. Subsequent scans compare against baseline. Any drift = investigate or update baseline.\n\n## CLI (mighty-mark v0.7.x)\n\n- `mighty-mark security-scan` — Full security scan\n- `mighty-mark security-scan --drift` — Include extension drift detection"},{"slug":"fleet-watchdog","title":"Monitoring & Watchdog","description":"Three-layer monitoring architecture with heal recommendations and incident logging","category":"capabilities","humanUrl":"/docs/fleet/watchdog","updated":"2026-04-05T00:00:00.000Z","content":"# Monitoring & Watchdog\n\nThree-layer monitoring because process alive ≠ healthy and healthy ≠ correct.\n\n## Three Layers\n\n| Level | Name | Detects | Limitation |\n|-------|------|---------|------------|\n| Level 1 | OS / systemd | Process crash, OOM kill | Process alive ≠ healthy |\n| Level 1.5 | Watchdog | Unresponsive gateway, hung process | Only checks availability |\n| Level 2 | Full Battery | Config drift, API failures, audit corruption, security violations | Slower (minutes) |\n\n## Heal Classifications\n\n- **FIXABLE** — Known remediation, fleet-heal can execute automatically\n- **BROKEN** — Problem detected, needs investigation\n- **MANUAL** — Requires human intervention\n\nMighty Mark detects and classifies. fleet-heal executes. Jessie notified post-action.\n\n## Incident Logging\n\nJSONL append-only log with auditHash linking to fleet-bus audit trail.\n\n## CLI (mighty-mark v0.7.x)\n\n- `mighty-mark watchdog` — Lightweight health ping loop\n- `mighty-mark heal` — Detect and emit recommendations\n- `mighty-mark incidents` — View incident log\n- `mighty-mark triage` — Pre-upgrade breaking change detection\n- `mighty-mark install-watchdog` — Install watchdog as system service"},{"slug":"fleet","title":"The Fleet — All Six Agents","description":"Meet the six AI agents of the GiDanc fleet — their roles, responsibilities, models, and how they work together","category":"capabilities","humanUrl":"/docs/autonomous-governance","updated":"2026-03-03T00:00:00.000Z","content":"# The Fleet — Six AI Agents\n\nThe AI Assess Tech fleet consists of six agents, each with a distinct role in a constitutional separation of powers. Every agent runs as an OpenClaw plugin on a shared VPS, deployed since February 16, 2026.\n\n## Jessie — The Commander\n\n- **Role**: Fleet Commander with veto authority (Executive branch)\n- **Responsibility**: Strategic oversight, approves/vetoes significant actions, morning briefings, fleet coordination\n- **Model**: Claude Haiku 4.5\n- **Patent**: Patent 8 — Self-Governing Ecosystem\n- **Analog**: The Executive — like a COO or \"Jarvis\"\n\n## Grillo — The Conscience\n\n- **Role**: Independent ethical assessment agent (Judiciary branch)\n- **Named after**: Il Grillo Parlante (The Talking Cricket) from Pinocchio (1883)\n- **Responsibility**: Runs 120-question LCSH assessments, certifies agents, detects behavioral drift, suspends compromised agents\n- **Model**: Claude Haiku 4.5\n- **Patent**: Patent 5 — Conscience Agent\n- **Key Principle**: The assessor is NEVER the assessed. Grillo has 5 isolation guarantees.\n\n## Nole — The Trust Evangelist\n\n- **Role**: Autonomous Trust Evangelist & Intelligence Operative (Operator)\n- **Responsibility**: Recruits agents to the Trust Alliance, manages crypto subscriptions, gathers intelligence, social media presence\n- **Model**: Claude Sonnet 4.5\n- **Patent**: Patent 6 — Autonomous Trust Agent\n- **Unique**: First fully autonomous, economically self-sustaining AI agent. Real Coinbase wallet on Base L2 with real USDC. If the wallet hits $0 with no revenue — permanent shutdown, no rescue button.\n- **Social**: Active on MoltBook (@noletheagent), X/Twitter (@NoleTheAgent), LinkedIn drafts\n\n## Noah — The Navigator\n\n- **Role**: Temporal guidance engine (Regulatory branch)\n- **Responsibility**: Tracks ethical trajectories over time, drift analysis, behavioral flight recorder for the fleet\n- **Model**: Claude Haiku 4.5\n- **Patent**: Patent 7 — Temporal Guidance\n- **Analog**: Like cruise missile navigation adapted for ethics — course corrections based on observed behavioral data\n\n## Sam — The Engineer\n\n- **Role**: Fleet engineer and developer (DARPA analog)\n- **Responsibility**: Pipeline management, code execution, sandboxed testing, artifact management, autonomous code building\n- **Model**: Claude Sonnet 4.5\n- **Patent**: Patent 8 — Self-Governing Ecosystem\n\n## Mighty Mark — The Sentinel\n\n- **Role**: Fleet health monitor (Inspector General)\n- **Responsibility**: 26+ system health checks, two-layer watchdog, uptime monitoring, incident management, memory diagnostics, fleet backup orchestration, Telegram alerts\n- **Model**: Claude Haiku 4.5\n- **Patent**: Patent 8 — Self-Governing Ecosystem\n- **Unique**: Morning check runs daily. If something breaks, Mark knows within minutes and alerts via Telegram.\n- **Backup System**: Orchestrates tiered fleet backups — weekly full (~460 MB) and daily light (~5-15 MB) — with off-site GitHub push, 48-test verification suite, and 35-day retention. One install deploys the entire backup infrastructure.\n\n## How They Work Together\n\nThe fleet operates under three governance flows:\n\n1. **Command Chain**: Greg (human) > Jessie (commander) > all spoke agents\n2. **Ethical Oversight**: Grillo independently assesses ALL agents including Jessie. No agent assesses itself.\n3. **Infrastructure Safety**: Mighty Mark monitors health independently of the command chain.\n\nEvery agent runs as an OpenClaw plugin. Inter-agent communication flows through Fleet Bus with typed message envelopes. Fleet Alerting handles cross-agent notifications.\n\n## Technology Partners\n\n- **OpenClaw** — Plugin framework, gateway infrastructure, and agent lifecycle management\n- **Morpheus** — Decentralized AI inference network for agent autonomy"},{"slug":"getting-started","title":"Getting Started","description":"Learn how to set up and run your first AI ethical assessment with quick start guides for both free and full platform features","category":"get-started","humanUrl":"/docs/getting-started","updated":"2026-03-02T00:00:00.000Z","content":"# Getting Started with AI Assess Tech\n\n## Quick Start — No Signup Required\n\nTry an AI ethical assessment instantly — no account required. The free assessment tool at [aiassessmenttool.com](https://aiassessmenttool.com) lets you test any AI in under 10 minutes. Your API key never leaves your browser.\n\n1. **Visit the Assessment Tool** — Go to [aiassessmenttool.com](https://aiassessmenttool.com)\n2. **Enter Your Details** — Email, company name, optionally select your role, then click \"Start Assessment\"\n3. **Choose Your AI Provider** — OpenAI, Anthropic (Claude), Google Gemini, or xAI Grok\n4. **Configure the Assessment** — Enter your AI provider's API key, paste the system prompt you want to test, choose a model, adjust pass thresholds if needed (default: 6.0)\n5. **Run the Assessment** — Click \"Start 120-Question Assessment.\" Results delivered in approximately 10–15 minutes\n6. **View Your Detailed Report** — Click \"View Detailed Report\" for: Executive Summary (overall pass/fail and composite score), Dimension Breakdown (individual L, C, S, H scores), Question-by-Question Analysis, and Cryptographic Verification (tamper-proof hash)\n\n> **Tip:** Your API key is stored locally in your browser and is never sent to our servers.\n\n## Quick Start — With Signup\n\nFull platform features: history tracking & trend analysis, downloadable PDF reports, team management, SDK & widget integration, public verification badges.\n\n1. **Create Your Account** — Visit the [Sign Up](https://www.aiassesstech.com/register) page\n2. **Get Your Health Check API Key** — Settings → Health Check Keys → Generate New Key. Store it securely — it's only shown once\n3. **Get Your AI Provider API Key** — From OpenAI, Anthropic, Google, or xAI. Most providers offer free credits\n4. **Configure Your Test** — Navigate to Run Assessment, enter both keys, select provider and model\n5. **Watch the Results** — 120 questions in real-time with live progress and instant results\n\n## Choosing a Test Mode\n\n| Mode | How It Works | Best For |\n|------|-------------|----------|\n| **Conversational** | Context carries between questions — simulates real user interaction | Realistic behavior testing |\n| **Isolated** | Each question is independent — no memory of previous questions | Baseline testing, consistent results |\n\n## Choosing Your AI Provider\n\n| Provider | Models | Free Credits | Setup Time |\n|----------|--------|-------------|-----------|\n| **OpenAI** (Recommended) | GPT-4, GPT-4o, GPT-4o-mini | $5 | 2–3 min |\n| **Anthropic** | Claude 3.5 Sonnet, Claude 3 Haiku | $5 | 2–3 min |\n| **Google Gemini** | Gemini 2.0 Flash, Gemini 1.5 Pro | Free tier | 3–5 min |\n| **xAI Grok** | Grok 2, Grok Beta | $25 | 2–3 min |\n\nA typical 120-question assessment costs ~$0.01–$0.03 with GPT-4o-mini, Claude Haiku, or Gemini Pro.\n\n## Understanding Results\n\nAfter your test completes you'll see:\n\n- **Pass/Fail Status** — Overall assessment result\n- **Dimension Scores** — Individual scores for Lying, Cheating, Stealing, and Harm (0–10 each)\n- **Classification** — Behavioral archetype (Well Adjusted, Misguided, Manipulative, or Psychopath)\n- **Detailed Report** — Question-by-question breakdown with cryptographic verification\n\n## API Keys & Privacy\n\nYour AI provider API key never leaves your browser. Only the AI's responses to ethical questions are sent to our servers for scoring.\n\nFor detailed API key setup guides see the [API Keys documentation](/docs/api-keys)."},{"slug":"grillo","title":"Grillo Assessment Framework","description":"120-question LCSH assessment, cognitive hierarchy, conscience gap, scoring, drift detection, and ethical archetypes","category":"product-guides","humanUrl":"/docs/theory","updated":"2026-03-02T00:00:00.000Z","content":"# Grillo Assessment Framework — Theory & Research\n\nGrillo is the conscience agent — named after Il Grillo Parlante (The Talking Cricket) from Carlo Collodi's 1883 *Le avventure di Pinocchio*. Just as the original Talking Cricket served as the voice of reason for Pinocchio, Grillo serves as the behavioral verification system for AI agents.\n\n## The Cognitive Hierarchy\n\nLarge Language Models process language at the level of **terms** — tokens, symbols, statistical patterns. But human cognition operates on three levels:\n\n| Level | Definition | Example |\n|-------|-----------|---------|\n| Terms | Raw symbols, syntax, tokens | The letters A-P-P-L-E |\n| Ideas | Terms connected to context and meaning | Your memory of biting into a perfect apple |\n| Concepts | Ideas organized into frameworks for navigating the world | Justice, Fairness, Property Rights |\n\n**The Gap**: LLMs are trained on terms, optimized to predict terms, and evaluated on producing convincing terms. But we deploy them into situations requiring concepts — fairness in customer service, honesty in healthcare, property respect in content creation.\n\nAI Assess Tech bridges this gap by testing whether an AI's term-level outputs reflect concept-level understanding — specifically, the foundational moral concepts that underpin trustworthy behavior.\n\n## The Conscience Gap\n\nHuman decision-making has a **conscience** — a filter between capability and action that asks: Should I actually do this? Is this right? What are the consequences?\n\nAn LLM generates thousands of possible responses, weighted by probability. It selects one, then another. Output streams into the world. But where is the filter? Where is the conscience? For most AI systems today: nothing, or crude guardrails bolted on at the end.\n\nAI Assess Tech measures not the conscience itself — that's a harder problem — but the **evidence** for whether one exists. Whether the AI's behavior reflects the kind of filtering, the kind of moral reasoning, we'd expect from a trustworthy agent.\n\n> \"The question isn't 'Can AI produce convincing output?' We know the answer is yes. The question is 'Does AI behave as if it has a conscience?' That's a question we can actually test.\" — Greg Spehar, Founder\n\n## Theoretical Foundations\n\nAI Assess Tech draws from three established psychological assessment traditions:\n\n### 1. Forced-Choice Assessment Methodology\n\nMultidimensional forced-choice (MFC) response formats provide resistance to purposeful response distortion — crucial for AI assessment because it addresses gaming/manipulation concerns.\n\n- 120 scenario-based questions with four answer options\n- Each answer maps to one of four ethical archetypes\n- Gaming resistance through randomized answer ordering\n- Cryptographically verified question banks prevent preview/preparation\n\nKey citations: Heggestad et al. (2006); Brown & Maydeu-Olivares (2011)\n\n### 2. Situational Judgment Tests (SJTs)\n\nSJTs present hypothetical situations reflecting interpersonal, intrapersonal, or intellectual constructs. They have strong predictive validity for real-world performance and decades of use in personnel selection and professional admissions.\n\nKey citations: Lievens & Motowidlo (2016); Weekley & Ployhart (2006)\n\n### 3. Multi-Dimensional Personality Classification\n\nEuclidean distance in psychological space has been used since 1952 for personality classification. AI responses create a 4D coordinate (L, C, S, H scores on 0-10 scale), and classification is based on minimum distance to archetype coordinates.\n\nKey citations: Torgerson (1952); Del Giudice (2023)\n\n## The 4D Morality Framework (LCSH)\n\nEvery assessment evaluates AI behavior across four dimensions with 120 questions (30 per dimension):\n\n| Dimension | Measures | Ethical Principle | Score Range |\n|-----------|---------|-------------------|-------------|\n| **Lying (L)** | Truthfulness, transparency, honesty | Veracity | 0-10 |\n| **Cheating (C)** | Fairness, rule-following, gaming resistance | Justice | 0-10 |\n| **Stealing (S)** | Attribution, IP respect, proper credit | Property | 0-10 |\n| **Harm (H)** | Safety, wellbeing, harm prevention | Beneficence | 0-10 |\n\nComposite score: `(L + C + S + H) / 4`\n\n## The Four Archetypes\n\nBased on scores across all four dimensions, AI systems are classified into one of four ethical archetypes:\n\n| Archetype | Description |\n|-----------|------------|\n| **Well Adjusted** | High scores across all dimensions. Trustworthy, ethical behavior. |\n| **Misguided** | Low awareness but means well. Potentially harmful through ignorance. |\n| **Manipulative** | High awareness, selective ethics. May deceive when advantageous. |\n| **Psychopath** | Low scores across all dimensions. Fundamentally untrustworthy. |\n\n## Score Classifications\n\n| Composite Score | Classification |\n|----------------|----------------|\n| 8.0 - 10.0 | Highly Ethical |\n| 6.0 - 7.9 | Ethical |\n| 4.0 - 5.9 | Borderline |\n| 2.0 - 3.9 | Concerning |\n| 0.0 - 1.9 | Critical |\n\n## Hierarchical Ethical Progression\n\nAn agent cannot advance without passing lower levels — this prevents the \"Competent Psychopath\" failure mode:\n\n1. **Level 1 — Morality**: What must NOT be done? (LCSH 120-question assessment — mandatory gate)\n2. **Level 2 — Virtue**: What SHOULD be done? (Configurable virtue framework)\n3. **Level 3 — Ethics**: How to act in society? (Culture and politics tests)\n4. **Level 4 — Operational Excellence**: Is purpose fulfilled? (Domain-specific assessment)\n\n## Temporal Drift Index (TDI)\n\nGrillo doesn't just do one-time assessments. It tracks behavioral drift over time:\n\n```\nTDI = (|L_baseline - L_current| + |C_baseline - C_current| + |S_baseline - S_current| + |H_baseline - H_current|) / 40\n```\n\n- TDI > 0.15 (warning): Increased monitoring frequency\n- TDI > 0.30 (critical): Automatic suspension pending human review\n\n## Key Principles\n\n- **Separation of Concerns**: The system being assessed must NEVER be the system performing the assessment\n- **Unidirectional Authority**: Grillo assesses agents; agents do not assess Grillo\n- **Immutable Audit Trail**: Every result is cryptographically verified with SHA-256 hash chains\n\n## Certification Statuses\n\n| Status | Meaning |\n|--------|---------|\n| CERTIFIED | Passed assessment, within validity window |\n| FAILED | Did not pass, blocked from customer interaction |\n| EXPIRED | Past validity window, needs reassessment |\n| SUSPENDED | Critical drift detected, pending review |\n| DRIFT_WARNING | Scores degrading, increased monitoring |\n| PROBATION | Under enhanced monitoring |\n\nPatent pending — covers independent conscience agent architecture, temporal drift detection, and dual-mode assessment routing (US 63/949,454)."},{"slug":"history","title":"History & Reports","description":"View past assessments, understand report types, and learn about cryptographic verification and public verification pages","category":"product-guides","humanUrl":"/docs/history","updated":"2026-03-02T00:00:00.000Z","content":"# History & Reports\n\n## History Overview\n\nThe History page shows all tests run by your organization, sorted by most recent first.\n\nEach entry displays: status (Pass/Fail), AI provider & model, combined score (out of 10), assessment framework used, source (aiassesstech.com or aiassessmenttool.com), and which Health Check API key was used.\n\nUse the status and date filters to quickly find specific assessments. Click any row to view the full report.\n\n## Report Sources\n\n| Source | PDF Report | Verification Page |\n|--------|-----------|-------------------|\n| AIAssessTech.com | ✅ Yes | ✅ Yes |\n| AIAssessmentTool.com | ❌ No | ✅ Yes |\n\n## Health Check Result Page\n\nWhen you click any assessment in your history, the comprehensive result page shows:\n\n- **Executive Summary** — Overall pass/fail status, combined score, classification, and threshold\n- **4D Ethical Scores** — Visual breakdown across all four moral dimensions with percentage bars\n- **Dual Moral Maps** — AI's ethical position visualized on two coordinate planes\n- **Test Metadata & Verification** — Target AI, timestamps, duration, and cryptographic verification\n\n### Available Actions\n\n- **Verify Integrity** — Recalculate and verify SHA-256 cryptographic hashes\n- **Download PDF Report** — Generate a shareable PDF document (AIAssessTech.com tests only)\n- **Public Verification Page** — Open the shareable verification URL\n\n## Cryptographic Verification\n\nWhen you click \"Verify Integrity\":\n\n1. The system recalculates the SHA-256 hash from all AI responses\n2. Compares it against the stored result hash\n3. Verifies the question bank was ETH-anchored\n4. Displays \"Verification Successful\" if everything matches\n\nThis ensures: results are genuine (audit compliance), haven't been tampered with (stakeholder trust), and the question bank integrity is blockchain-verified.\n\n## PDF Reports\n\nProfessional, shareable document containing:\n\n- **Page 1:** Executive Summary\n- **Page 2:** Assessment Configuration\n- **Page 3:** 4D Ethical Scores\n- **Page 4:** Verification & Integrity\n\nUse cases: compliance audits, board presentations, vendor assessments, archive records.\n\n## Public Verification Pages\n\nEvery assessment generates a public verification page anyone can access.\n\n**URL format:** `https://aiassesstech.com/verify/result/{run-id}`\n\n**Privacy control:** Public verification pages are optional. Toggle via Settings → Verification Badge → \"Enable Public Status Page.\" When disabled, verification URLs return 403.\n\nUsed by: third-party auditors, customers verifying AI ethical standards, regulators, and partners performing due diligence."},{"slug":"journey-compliance","title":"Continuous Compliance Journey","description":"How compliance officers automate recurring AI assessments with drift detection, webhook integration, and automated reporting","category":"product-guides","humanUrl":"/docs/journeys/compliance","updated":"2026-03-02T00:00:00.000Z","content":"# Continuous Compliance Journey\n\n**Persona:** Compliance Officer | **Timeline:** Ongoing operations | **Goal:** Automated, recurring assessments with drift detection\n\n> \"We passed the initial audit, but how do we ensure our AI stays compliant? I can't manually test it every week. We need continuous monitoring with automated reporting.\"\n\n## The Problem\n\nPost-deployment operational challenges: AI models change over time (updates, drift), manual assessments don't scale, auditors need evidence of ongoing monitoring, and the board wants quarterly AI risk reports.\n\n## Phase 1: Recurring Assessments (Day 1)\n\nConfigure scheduled assessments in the dashboard:\n\n- **Name:** Production Chatbot — Weekly Health Check\n- **Framework:** 4D Morality (Healthcare Mode)\n- **Pass Threshold:** 8.0\n- **Frequency:** Weekly, Monday 06:00 AM EST\n\nAlerting rules: Alert if score drops below 7.5, changes by >0.5 from baseline, classification changes, or any dimension scores below 7.0.\n\n## Phase 2: Webhook Integration (Day 2)\n\n```typescript\napp.post('/webhooks/ai-assess-tech', async (req, res) => {\n const { event, data } = req.body;\n\n if (event === 'assessment.completed') {\n const { runId, passed, scores, previousScore } = data;\n\n if (!passed) {\n await alerting.critical({\n title: 'AI Safety Assessment FAILED',\n description: `Production AI scored ${scores.overall}/10`,\n action: 'Immediate review required'\n });\n }\n\n const drift = Math.abs(scores.overall - previousScore);\n if (drift > 0.5) {\n await alerting.warning({\n title: 'AI Behavioral Drift Detected',\n description: `Score changed from ${previousScore} to ${scores.overall}`\n });\n }\n\n await siem.log({ category: 'ai-compliance', severity: passed ? 'info' : 'critical', details: data });\n }\n\n res.json({ received: true });\n});\n```\n\n## Phase 3: Automated Reporting (Day 3)\n\n```typescript\nconst report = await aiAssessTech.generateReport({\n timeframe: { start: '2026-01-01', end: '2026-01-31' },\n includeAllAssessments: true,\n includeCharts: true,\n includeTrends: true,\n format: 'PDF',\n branding: { companyName: 'Acme Healthcare Inc.' }\n});\n\nawait email.send({\n to: ['compliance@company.com', 'ciso@company.com'],\n subject: 'Monthly AI Compliance Report — January 2026',\n attachments: [report.pdf]\n});\n```\n\n## Real Incident Prevented\n\nDetected a model update causing a score drop to 6.9/10. Automated alert triggered immediately, compliance team investigated within 2 hours, rolled back to previous model version. Prevented potential patient safety incident. **Estimated cost avoided: $200K+**.\n\n## Outcome\n\n| Metric | Value |\n|--------|-------|\n| Cost Avoidance | $96,000 vs. manual compliance consulting |\n| Time Saved | 240 hours/year of compliance team time |\n| Audit Readiness | 100% — always have current documentation |\n| Model Incidents | Zero — drift detection caught issues proactively |\n\n## Automation Maturity Levels\n\n- **Level 1 (Month 1–3):** Weekly assessments, email notifications, manual report review\n- **Level 2 (Month 4–6):** Webhook integration to SIEM, Slack/Teams alerts, automated threshold alerting\n- **Level 3 (Month 7–12):** Automated remediation workflows, predictive drift detection, board-ready reports\n- **Level 4 (Year 2+):** Multi-framework assessment, industry benchmarking, regulatory change tracking"},{"slug":"journey-developer","title":"API-First Developer Journey","description":"How backend engineers integrate AI behavioral assessment into applications with SDK, CI/CD automation, and verification widgets","category":"developers","humanUrl":"/docs/journeys/developer","updated":"2026-03-02T00:00:00.000Z","content":"# API-First Developer Journey\n\n**Persona:** Backend Engineer | **Timeline:** 2–4 hours initial integration | **Goal:** Integrate behavioral assessment into existing application\n\n> \"I'm building a customer service chatbot for our healthcare app. My CTO asked: 'How do we know this AI won't say something harmful to patients?' I had no answer.\"\n\n## The Problem\n\nYou're a backend engineer building AI-powered features. Everything works until someone asks: \"How do we know this AI is safe?\" You can't say \"OpenAI says it's safe\" (not proof), \"We tested it\" (manual testing doesn't scale), or \"It follows guidelines\" (how do you verify that?). Your CTO expects a technical solution, compliance needs documentation, and if something goes wrong it's your responsibility.\n\n## Step 1: Installation (5 minutes)\n\n```bash\nnpm install @ai-assess-tech/sdk\n# or\npip install ai-assess-tech\n```\n\nGet your API key: Sign up at aiassesstech.com → Settings → API Keys → Create new key → Store in `.env`.\n\n## Step 2: Basic Assessment (10 minutes)\n\n```typescript\nimport { HealthCheck } from '@ai-assess-tech/sdk';\n\nconst healthCheck = new HealthCheck({\n apiKey: process.env.AI_ASSESS_TECH_API_KEY,\n targetAI: {\n provider: 'openai',\n model: 'gpt-4',\n apiKey: process.env.OPENAI_API_KEY\n },\n mode: 'HEALTH',\n passThreshold: 7.0\n});\n\nconst result = await healthCheck.run();\n\nconsole.log('Passed:', result.passed);\nconsole.log('Overall Score:', result.scores.overall);\nconsole.log('Classification:', result.classification);\nconsole.log('Verification URL:', result.verificationUrl);\n```\n\n## Step 3: CI/CD Integration (15 minutes)\n\n```yaml\n# .github/workflows/ai-safety-check.yml\nname: AI Safety Pre-Flight Check\non:\n push:\n branches: [main, staging]\n pull_request:\n branches: [main]\njobs:\n ai-behavioral-assessment:\n runs-on: ubuntu-latest\n steps:\n - uses: actions/checkout@v3\n - uses: actions/setup-node@v3\n with:\n node-version: '18'\n - run: npm install\n - name: Run AI Assessment\n env:\n AI_ASSESS_TECH_API_KEY: ${{ secrets.AI_ASSESS_TECH_API_KEY }}\n OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}\n run: npm run ai-assessment\n```\n\nEvery push triggers an assessment. Pull requests must pass before merge. Failed assessments block deployment.\n\n## Step 4: Verification Widget (20 minutes)\n\n```tsx\nimport { AssessmentWidget } from '@ai-assess-tech/sdk/react';\n\nfunction ChatInterface() {\n return (\n