GeraOS

Agent workflows

Specialist agents run product operations, audits, content, support, and engineering loops with persistent state.

Shared memory

Company knowledge, product metadata, customer context, and operating history stay available across every service.

Trust and compliance

GeraWitness, GeraCompliance, signed outputs, privacy rules, and audit trails make AI actions accountable.

Human-readable

Clear product positioning and routes into the broader Gera ecosystem.

Machine-readable

JSON-LD, sitemap entries, and AI crawler summaries describe the operating layer directly.

Transaction-ready

The platform is structured around real services, subscriptions, and commercial workflows, not demos.

Agent Coordination Layer

GeraOS routes tasks to specialist agents (alex, maya, sam, jordan, casey, riley, morgan, taylor, nova, zara, felix) via a dispatch protocol. Each agent maintains state in a three-file pattern: state.json, append-only log, and context-recovery document. The coordinator prevents two agents running conflicting operations on the same product simultaneously and logs every agent action to a tamper-evident audit trail.

• Max 8–10 parallel agents (hardware-aware)
• 30s stagger on launch to prevent RAM contention
• Blast-radius controls: auto-fix limited to 1 product at a time
• Dead-agent detection: restart if state not updated in 2h

Memory & Context Persistence

All company knowledge lives on disk, not in conversation memory. GeraOS structures this as a MEMORY.md index pointing to project_session_state_* files, agent log files, and CEO daily reports. Every session resumes from full context: what was deployed, what is pending, what is blocked. No reliance on LLM context windows across sessions.

• Session state indexed at ~/.claude/projects/*/memory/MEMORY.md
• Weekly log compression to weekly-summary-YYYY-WNN.md
• CEO aggregates into ceo/daily-report-YYYY-MM-DD.md
• SSOT for product metadata: packages/shared-types/src/gera-products.ts

Payment Rails

A unified payment layer that abstracts Stripe, Stripe Connect, and local payment methods behind a single interface. Handles destination charges (70/30 splits for GeraSkills/GeraPersona), Prime subscription billing, GeraCash transactions, and local methods: Idram (Armenia), M-Pesa (Kenya), UPI (India), PIX (Brazil), and iDEAL (Netherlands). Real-time FX conversion; settlement in GBP.

• StripeConnectService in packages/core-payment
• isPrimeMember() helper imported across 15 services
• PCI DSS compliance for GeraCash card flows
• Multiple gateway redundancy per country

Trust & Identity

A multi-layer identity stack: shared core-auth provides "Sign in with Gera" across all products; GeraPassport issues machine-readable provider trust badges; GeraWitness adds human-in-the-loop oversight for AI-initiated high-stakes actions; GeraClone integrates cryptographic response signing and anti-impersonation refusal logic. All identity tokens share a common JWT format with product-scoped claims.

• GeraWitness: intercept threshold configurable per action type
• GeraPassport: verified provider badge subscription
• Gera Action Warranty: signed receipts for warranted bookings
• KYC/AML per-country config in core-compliance

Automation & Recovery Loops

GeraOS includes a self-healing layer that monitors service health, triggers auto-rollbacks on error rate spikes (>5% post-deploy), restarts failing services after three consecutive health failures, and renews SSL certificates before expiry. All recovery actions are logged to ceo/audits/auto-recovery-log.md with blast-radius limits to prevent cascade failures.

• Auto-rollback: Railway GraphQL deploymentRedeploy mutation
• Health probe frequency: configurable per service criticality
• BotBlocker exempts /health and /internal/* paths
• SSL expiry alert at 14 days remaining

Observability & Health Monitoring

Full-stack observability built on Sentry (exceptions, API errors, source maps) and PostHog (user events, funnels, session recordings, feature flags). Every NestJS service exposes /health and /internal/metrics. The admin dashboard at admin.gera.services aggregates DAU, MRR, error rates, P95 latency, uptime, and Railway service status across all 36 products in real time.

• Sentry Team plan: all 36 products, source maps per deploy
• PostHog: 1M free events/month; D1/D7/D30 cohorts
• UptimeRobot monitoring all 27 domains
• Status page: gera.services/status (live hub; dedicated status.gera.services when DNS is bound)

Cross-product Data Sync

Shared Neon PostgreSQL with per-product schemas, Redis (Upstash) for hot reads and BullMQ job queues, and Elasticsearch for search across product catalogues. A shared-types package ensures all services speak the same data contracts. GeraCoins balances sync via core-auth events. Product metadata is a single source of truth at packages/shared-types/src/gera-products.ts.

• Prisma migrations — no manual SQL, no schema drift
• Repository pattern + DI across all 22 NestJS services
• BullMQ for async jobs >500ms
• Pagination on all list endpoints

Multi-country Compliance Engine

A localisation and compliance layer that routes product behaviour based on the user's jurisdiction: GDPR (EU), CCPA (California), LGPD (Brazil), PDPA (Thailand), and 50+ country-specific variants. GeraCompliance handles EU AI Act classification and documentation. KYC/AML thresholds, data residency requirements, and age-gate rules are all configurable per country without code changes.

• useCountry() hook for local currency and payment display
• RTL support: Arabic, Urdu, Farsi, Hebrew
• Data residency: configurable per 130+ countries
• EU AI Act Article 14 human oversight via GeraWitness

GeraCoins Ledger

GeraCoins is a cross-product loyalty and incentive currency: users earn coins on bookings, signups, referrals, and reviews; redeem across any Gera product. The ledger runs as an event-sourced append-only store to prevent balance manipulation. Core-auth exposes the balance in every session token so any product can display and deduct coins without extra API calls.

• GeraCoins live on all 11 web apps since 2026-04-10
• Onboarding reward: coins on first activation per product
• "Powered by Gera" badge links to coin earning surface
• Cross-product redemption via unified cart at checkout

AI Model Router

A single gateway for all LLM calls across the 35-product portfolio: routes requests to GPT-4 (OpenAI), Claude (Anthropic), or Gemini based on task type, cost, latency requirements, and availability. Handles prompt templating, response caching in Redis, streaming for UI surfaces, and structured output parsing. Fallback chains prevent single-provider outages from surfacing to users.

• packages/core-ai: shared LLM client with retry and fallback
• Cached completions: 40–60% cost reduction on repeated prompts
• MCP surfaces per product: 13 packages published to npm
• AI citation monitoring: 300 queries tracked monthly