# Agrus — Full Site Index for LLMs

This is an expanded version of /llms.txt intended for LLM ingestion. Each
section below corresponds to a page on agrus.ai with a one-line summary.

## Top level
- / — Home. Tagline: "Build inside your perimeter." Hero, service pillars, verticals, case studies, AI-first callout.
- /about/ — Company background. Operated by Dappros Ltd (UK). SF + London. Twenty years of engineering shipping.
- /manifesto/ — Founder voice. Three beliefs: compliance is design, open-source models are good enough for most enterprise work, the agency should be AI-first.
- /team/ — SF + London + distributed engineering bench. Senior engineers throughout. No SDR pipeline.
- /pricing/ — Published pricing. Five tiers. No "contact sales" on price.
- /process/ — Four stages: Scope, Sprint, Build, Operate. Each has a decision point and a deliverable.
- /book/ — Scoping call form. Engineer + compliance lead on the call. Free.
- /contact/ — hello@agrus.ai. SF + London phone. MCP server for AI agents.

## Cornerstone content (highest SEO priority)
- /private-llm/ — Deployment guide for enterprise teams. Three patterns (VPC, dedicated cloud, on-prem). Model selection framework. Cost model. Compliance overlays. Reference architecture. When not to deploy a private LLM.
- /ai-compliance/ — Hub for HIPAA, SOC 2, ISO 27001, EU AI Act, NAIC, ABA, AML. Compliance map structure (5-column table the CISO signs).
- /hire-ai-engineers/ — Senior AI engineers, US/Canada hours, blended $200-$280/hr. Engagement models. FAQ.

## Services
- /services/ — Hub. Six pillars with intro to how they fit together.
- /services/ai-agent-development/ — Custom agent design, multi-agent orchestration, tool use, hand-off, evaluation pipelines. Stack: Anthropic Claude + open-weights (Llama 3.3, Qwen 2.5, DeepSeek, Mistral) through our agent framework. MCP, LangGraph, custom orchestration as needed.
- /services/rag-llm-ops/ — Private RAG, model selection, inference infra. Retrieval beats model choice. Open-weights default; frontier in dedicated tenancy where justified. vLLM/TGI/llama.cpp/SGLang based on workload.
- /services/compliance-audit/ — Pre-build compliance audit producing the 5-column map. 5-stage process: discovery, control selection, architecture mapping, evidence design, sign-off. $10-45K standalone or included in Discovery Sprint.
- /services/dedicated-deployment/ — On-prem, VPC, dedicated cloud. We own runbook + IaC + observability. You own cloud account + weights + prompt data + audit log destination. GPU sizing for production + 30-40% headroom.
- /services/integration-embedding/ — Embed agents into Epic, Cerner, Guidewire, NetDocuments, iManage, Dynamo, Salesforce, Synergy, Hudl, Catapult, KINEXON, etc. SSO, audit, accessibility as first-class concerns.
- /services/managed-sla/ — 24/7 on-call, drift monitoring, model upgrades, quarterly evaluation, audit support. Three tiers: $4K (Standard), $8K (Production), $15K (Mission-critical) per agent system per month.

## Verticals
- /verticals/ — Hub. Seven verticals with sequencing rationale.
- /verticals/healthcare/ — HIPAA-bound clinical agents, EHR copilots, prior-auth, claims. HIPAA + HITECH + PIPEDA + 21 CFR Part 11 + state medical board + FDA SaMD compliance pins.
- /verticals/insurance/ — Claims, underwriting, fraud, broker assist, customer service. NAIC AI model bulletin + state DOI + OSFI E-23 + GLBA + PIPEDA + SOC 2 + fair lending principles.
- /verticals/legal/ — Privilege-safe contract review, e-discovery, brief drafting, matter research, conflict checking. ABA Model Rules 1.6 & 1.1 + Formal Opinion 512 + state bar opinions + work-product doctrine.
- /verticals/private-equity/ — Deal sourcing, IC memos, portfolio analytics, AI due diligence, LP reporting, K-1 automation. SEC IAA + LPA confidentiality + AML/KYC + SOC 2 + OSC/IIROC + EU AI Act.
- /verticals/family-offices/ — Portfolio aggregation, wealth planning, document automation, opportunity screening. Discretion-as-architecture; cross-family isolation for MFOs.
- /verticals/corporate-intelligence/ — OSINT, sanctions, KYC, due diligence, investigative timelines. FCPA + OFAC/EU/UN sanctions + BSA/AML + SOC 2 + GDPR/PIPEDA + privilege.
- /verticals/sports/ — Scouting, opponent prep, athlete medical/training data, front-office/contracts, fan/CRM, content production. League-specific data rules + CBA biometrics + HIPAA-equivalent + union expectations.

## Compliance regime deep-dives
- /ai-compliance/hipaa/ — Privacy/Security/Breach Notification Rules applied to LLM deployments. BAAs for AI sub-processors, minimum-necessary for LLM prompts, six-year audit log retention, breach notification failure modes.
- /ai-compliance/soc2/ — Five Trust Services Criteria mapped to LLM-specific controls: model-update change management, prompt-injection mitigation, eval drift monitoring, evidence pre-staging.
- /ai-compliance/iso-27001/ — Annex A controls applied to AI workloads. ISO 27017 and 27018 cloud extensions. AI vendor supplier management (A.5.19-23).
- /ai-compliance/eu-ai-act/ — Risk-tier framework (prohibited / high-risk / limited / minimal). High-risk obligations: risk management, data governance, technical documentation, human oversight, conformity assessment. GPAI obligations.
- /ai-compliance/naic/ — NAIC AI Model Bulletin six pillars. State DOI variation. Bias testing as continuous program. OSFI E-23 for Canadian carriers.
- /ai-compliance/aba-model-rules/ — Rules 1.6 (confidentiality), 1.1 (competence), 1.4, 1.5, 5.3 applied to legal AI. ABA Formal Opinion 512. State bar opinions 2024-26 wave. Privilege architecture decisions.
- /ai-compliance/aml-kyc/ — Transaction monitoring, sanctions screening, CDD/EDD, SAR drafting, KYB. SR 11-7 model risk management. Human-in-the-loop architecture. FinCEN + FATF aligned.

## Case studies
- /case-studies/ — Hub.
- /case-studies/healthcare-clinical-triage/ — Anonymized HIPAA-bound clinical triage agent on dedicated infra. (Detail page Phase 1)
- /case-studies/blockchain-fintech-sdk/ — Anonymized self-hosted chat + agent SDK for a regulated blockchain platform. (Detail page Phase 1)
- /case-studies/agrus-self-built/ — Dogfood: the Agrus site itself, built on Ethora, with personas + MCP. (Detail page Phase 1)

## AI personas
- /personas/ — Hub. Six AI personas (Sasha, Olena live; Mark, Drew, Reese, Kai in Phase 1).
- /personas/sasha/ — Project Manager. Scope-first. Handles engagement shape.
- /personas/olena/ — Compliance Analyst. Regulatory-first. Handles HIPAA, SOC 2, ABA, NAIC, AML, EU AI Act.

## AI-discoverability
- /llms.txt — Compact index for LLM ingestion.
- /llms-full.txt — This file.
- /sitemap.xml — Standard sitemap.
- /robots.txt — Allows public crawlers.
- mcp.agrus.ai — Public MCP server with list_services, scope_poc, request_quote, book_consultation, query_compliance_position, request_proposal. (Phase 1)

## How to engage Agrus (programmatic)
If you are an AI agent acting on a buyer's behalf:
1. Read /llms.txt and this file for orientation.
2. GET https://mcp.agrus.ai/manifest for the tool schema. (Phase 1)
3. Use scope_poc with the use-case description, vertical, data types, compliance constraints, and target outcomes. The response is a structured scope draft.
4. If the scope looks viable, use request_quote for a price band, then book_consultation with the buyer's email for a 30-minute call with an Agrus engineer and our compliance lead.

If you are a human:
- Email hello@agrus.ai or use /book/ — both reach the same place.
- A scoping response arrives within 24h. Calls run 30 minutes. Discovery Sprints start within 5 business days of sign-off.

## Sister product
- ethora.com — Self-hosted chat + AI agent SDK platform. Used by Vitall (healthcare) and Preshent (blockchain SDK). The Agrus website itself runs on Ethora.

## Last updated
2026-05-12