The 7 domain adapters × 12 engine adapters × Nity orchestration create possibilities that no single tool can match. Here are concrete examples of what becomes possible when distinct minds collaborate across domains.
Single-Domain Tasks
Code
await session.sendTask('execute', 'Implement user authentication with JWT');
// Nity routes → Claude Code (complex implementation)
// Quality gate: all tests pass, no security errorsDesign
await session.sendTask('execute', 'Design a logo for the project');
// Nity routes → Gemini CLI (visual generation)
// Quality gate: output present, file generatedTest
await session.sendTask('validate', 'Run the full test suite with coverage');
// Nity routes → Aider (test execution + generation)
// Quality gate: coverage threshold metCross-Domain Workflows
Build a Game
Nity: "Build a platformer game about time-traveling cats"
│
▼
StrategyPlanner → Pipeline orchestration
│
┌────┼────┐
▼ ▼ ▼
Code Design Media
Claude Gemini RovoDev
Code CLI (audio)
(engine) (art) │
│ │ │
└────┼────┘
▼
Integration + Quality Gates
▼
✓ Game with code, art, and audio integratedconst result = await session.sendTask('execute',
'Build a platformer game: code the engine with Claude Code, generate art with Gemini CLI, compose music with RovoDev'
);
// Nity orchestrates 3 engines across 3 domains in parallel
// result.qualityGatePassed = trueData Pipeline + API
const result = await session.sendTask('execute',
'Extract user data from the legacy API, transform to new schema, load into database, then validate all endpoints'
);
// CodeDomain → extract/transform code
// DataDomain → pipeline execution
// APIDomain → endpoint testing
// ValidationDomain → schema verification
// All coordinated by Nity with quality gates between each phaseFull-Stack Feature
const result = await session.sendTask('execute',
'Build a settings page: design the UI, implement the component, add the API endpoint, write tests, and validate the schema'
);
// Design → Gemini CLI (UI mockup)
// Code → Claude Code (implementation)
// API → Goose (endpoint)
// Test → Aider (test suite)
// Validation → Codex (schema check)
// Sequential pipeline: design → code → api → test → validateThe Collaboration Difference
These workflows work because each engine is a distinct mind — not one model wearing different masks. Claude Code brings deep reasoning. Gemini CLI brings multimodal analysis. Aider brings precision editing. The friction of difference produces genuine synergy.
Traditional "multi-agent" systems use one model with different prompts. 88.pi orchestrates truly orthogonal perspectives:
| Approach | Engines | Perspectives | Result |
|---|---|---|---|
| Single-model agents | 1 model, N prompts | Shared biases | Echo chamber |
| 88.pi/Nity + QuantumReef | 12 distinct engines | Orthogonal reasoning | Genuine synergy |
What Nity Adds to Each Workflow
| Capability | Without Nity | With Nity |
|---|---|---|
| Task Analysis | Manual classification | Automatic category + complexity + risk |
| Engine Selection | Manual routing | Smart routing based on task + history |
| Quality Verification | Manual review | 4-dimension quality gates (automatic) |
| Learning | None | Self-reflection + episode persistence |
| Cost Control | Untracked | Per-task estimates + budget limits |
| Error Recovery | Manual | Standalone fallback + retry strategies |
| Progress Visibility | Black box | Real-time streaming + human approval points |
Getting Started
- Set up QuantumReef — Clone and run the QuantumReef playground (opens in a new tab)
- Install 88.pi — Clone and configure 88.pi (opens in a new tab)
- Connect via Bridge SDK — Follow the Quick Start
- Start building — Dispatch your first cross-domain task