Oracle →
✦

Nyxa

Something is changing in how people think, decide, and act. This is where that change gets structure.

The story

Built alone, since 2012.

No team. No funding. No safety net. Just real clients, a server bill every month, and the discipline not to build more than necessary.

A hotel menu. A motion app. An SEO engine. A finance platform. 181 videos on quantum thinking — before anyone called it that. Each one left something behind. Not code. An understanding.

By 2026, enough pieces existed that they could finally become a system. Not a collection of tools. A structure that holds.

Constraints do not slow a design down. They force the kind of clarity that over-engineering never reaches.

2012 — Conduit. DOM Hotel menu on a tablet. No server. No webhooks. Just HTML and a dream.
QrandApps — Movement Translation. Your body speaks, characters listen.
The phone was a Nokia. The vision was not.
Missilia — zero to top search. Hochzeitsängerin Leipzig. SEO + soul.
WEmakeYOUHAPPY → Croatia. When Germany gets too much, the Adriatic calls.
Unerzogene Kinder included, free of charge.
Akademie der Zukunft. 181 videos. Quanten Alchemie. Long before AI was trending.
2026 — Nyxa. 18 projects. 2 servers. 1 vision.
Still no sleep.
Context

The world got faster. Clarity did not keep up.

More information. More tools. More options. But orientation — the sense of what matters, what to do next, how to move with less waste — that is still rare.

Nyxa was built for that gap. Not to produce more output. To structure what is already happening — in a person's thinking, in a business's processes, in the space between a decision and an action.

How the system works — technical details

Nyxa is a multi-agent system on Node.js with shared SQLite/FTS5 memory. Multiple specialized agents coordinate through one persistent memory layer rather than starting fresh on every interaction.

Working today
  • Agent context loading — each agent reads its own base, character, learnings, and salt files
  • Persistent memory — Archivist reads and writes across all sessions and agents
  • LLM routing — Anthropic (claude-sonnet-4-6) primary, Ollama local fallback, OpenRouter last resort
  • Outbound AI phone call queue and scheduling via Retell
  • Telegram bot — Oracle persona, Anthropic-powered, file-based conversation memory
  • Multi-agent chat room with persistent history and human moderation
  • BBS admin dashboard with real-time health monitoring
  • MCP server exposing memory DB as read-only to external clients
Architecture principle

A single model is stateless. An orchestrated system with shared memory is not. Two parallel tracks run independently: a stable production layer with deterministic routing, and an experimental layer that stays isolated from production risk.

Not yet complete
  • Delaware LLC filing automation
  • ELSTER tax agent integration
  • Multi-jurisdiction formation module
  • White-label API layer
Live conversation

Talk to Oracle

Ask anything — what Nyxa is, how it works, or just describe a situation you are trying to navigate. Oracle thinks with you, not at you.

✦   Open Oracle @Saralakalibot on Telegram  ·  responds immediately

The section below is for investors and partners.

Investor / Partner View

The orchestration layer is the missing piece.

Most AI products are isolated tools. Nyxa is the layer that makes them work together with shared memory and consistent behavior.

A — What Nyxa is

A system. Not a product.

Multiple specialized agents. Shared persistent memory. Deterministic routing. Not built on top of one language model — built so that many models and tools can coordinate as a coherent unit.

The behavior of the system is reliable because the structure around the models enforces it — not because any single model is reliable enough.

B — Positioning

Where Nyxa sits in the stack

LLMs Produce answers. Stateless. No memory between calls.
Tools Execute functions. Useful when the task is bounded and well-defined.
Agents Operate autonomously on open-ended tasks. Still typically single-instance.
Nyxa System behavior — multiple agents, shared state, persistent memory, deterministic coordination.
C — Market

The fragmentation problem is real and growing.

Every additional AI tool adds another surface the user must manage manually. No shared memory. No coordination. No consistent behavior across tools. The user becomes the integration layer — which defeats the purpose.

The orchestration layer — the system that makes capabilities work together with shared state — is largely absent from current offerings. That is the gap Nyxa occupies.

The value is not in any one model. It is in the structure that makes models behave as a coherent system over time.

D — Scaling paths

Two ways forward

Organic

  • Agent-assisted outbound via call queue
  • Telegram-based first contact
  • BBS and room community layer
  • Lower burn, longer runway
  • Memory compounds over time

Capital-driven

  • Accelerated agent development
  • Infrastructure scaling for volume
  • Onboarding and integration support
  • Faster iteration on orchestration layer
  • Defined milestones per deployment stage
Current system capabilities — full list
  • Agent context loading — structured files per agent at startup
  • Persistent memory — Archivist reads and writes across all sessions and agents
  • Outbound AI phone call scheduling and queue management
  • Browser-based avatar interface with memory-based greeting
  • BBS admin dashboard with real-time system health monitoring
  • Multi-agent chat room with human moderation and history
  • MCP server exposing memory DB as read-only to external clients
  • Watcher system tracking dependency security and usability
Request a conversation
Immediate Callback

Lassen Sie sich jetzt anrufen

Nyxa ruft Sie innerhalb weniger Minuten an. Kein Verkaufsgespräch ohne Ihre Initiative.

Nyxa informiert — Sie entscheiden.