|
|
--- |
|
|
license: mit |
|
|
library_name: hct-spec |
|
|
tags: |
|
|
- multi-agent |
|
|
- coordination |
|
|
- agent-orchestration |
|
|
- llm-agents |
|
|
- mcp |
|
|
- a2a |
|
|
--- |
|
|
|
|
|
# Harmonic Coordination Theory (HCT) - Specification |
|
|
|
|
|
**The canonical specification for HCT coordination signals and performance parameters.** |
|
|
|
|
|
## Overview |
|
|
|
|
|
Harmonic Coordination Theory (HCT) proposes a musical ontology for multi-agent coordination. Current frameworks (LangGraph, CrewAI, AutoGen) give you orchestration tools but lack a shared language for **coordination semantics**—timing, quality, intent, and harmony. |
|
|
|
|
|
HCT fills this gap using musical performance as the ontology: cues, fermatas, tempo, and dissonance become engineering primitives. |
|
|
|
|
|
## The 7 Coordination Signals |
|
|
|
|
|
| Signal | Meaning | Use Case | Musical Origin | |
|
|
|--------|---------|----------|----------------| |
|
|
| **CUE** | "Your turn—act now" | Task dispatch | Conductor's downbeat | |
|
|
| **FERMATA** | "Hold for approval" | Human-in-the-loop | Sustained note | |
|
|
| **ATTACCA** | "Immediate handoff" | Real-time flows | Seamless movement transition | |
|
|
| **VAMP** | "Loop until ready" | Quality checks | Repeat until cue | |
|
|
| **CAESURA** | "Full stop" | Emergency shutdown | Dramatic silence | |
|
|
| **TACET** | "Stay silent" | Resource conservation | Instrument rests | |
|
|
| **DOWNBEAT** | "Sync point" | Barrier synchronization | Unified entry | |
|
|
|
|
|
## Quick Start |
|
|
|
|
|
### Python |
|
|
```python |
|
|
from hct_mcp_signals import cue, fermata |
|
|
|
|
|
# Signal the analyst to start with high urgency |
|
|
signal = cue("orchestrator", ["analyst"], urgency=9, tempo="presto") |
|
|
|
|
|
# Hold for human approval before publishing |
|
|
hold = fermata("report_agent", "Ready for compliance review", hold_type="human") |
|
|
``` |
|
|
|
|
|
### TypeScript |
|
|
```typescript |
|
|
import { cue, fermata } from '@hct-mcp/signals'; |
|
|
|
|
|
const signal = cue("orchestrator", ["analyst"], { urgency: 9, tempo: "presto" }); |
|
|
const hold = fermata("report_agent", "Ready for review", { holdType: "human" }); |
|
|
``` |
|
|
|
|
|
### Rust |
|
|
```rust |
|
|
use hct_mcp_signals::{cue, fermata, Urgency}; |
|
|
|
|
|
let signal = cue("orchestrator", &["analyst"], Urgency::Nine, Tempo::Presto); |
|
|
let hold = fermata("report_agent", "Ready for review", HoldType::Human); |
|
|
``` |
|
|
|
|
|
### Go |
|
|
```go |
|
|
import "github.com/stefanwiest/hct-mcp-signals/go" |
|
|
|
|
|
signal := cue.Cue("orchestrator", []string{"analyst"}, 9, "presto") |
|
|
hold := fermata.Fermata("report_agent", "Ready for review", "human") |
|
|
``` |
|
|
|
|
|
## Installation |
|
|
|
|
|
```bash |
|
|
# Python |
|
|
pip install hct-mcp-signals |
|
|
|
|
|
# Node.js |
|
|
npm install @hct-mcp/signals |
|
|
|
|
|
# Rust |
|
|
cargo add hct-mcp-signals |
|
|
|
|
|
# Go |
|
|
go get github.com/stefanwiest/hct-mcp-signals/go |
|
|
``` |
|
|
|
|
|
## Protocol Extensions |
|
|
|
|
|
### HCT-MCP Signals |
|
|
**For Anthropic MCP Protocol** — The coordination layer that MCP is missing. |
|
|
|
|
|
[](https://pypi.org/project/hct-mcp-signals/) |
|
|
[](https://www.npmjs.com/package/@hct-mcp/signals) |
|
|
[](https://crates.io/crates/hct-mcp-signals) |
|
|
|
|
|
→ [GitHub: stefanwiest/hct-mcp-signals](https://github.com/stefanwiest/hct-mcp-signals) |
|
|
|
|
|
### HCT-A2A Extension |
|
|
**For Google A2A Protocol** — Coordination semantics for decentralized agent meshes. |
|
|
|
|
|
→ [GitHub: stefanwiest/hct-a2a](https://github.com/stefanwiest/hct-a2a) |
|
|
|
|
|
## Research Papers |
|
|
|
|
|
- [Harmonic Coordination Theory](https://www.stefanwiest.de/research) — CS.AI / CS.MA publications |
|
|
- [The 97% Problem](https://seekingsota.com/p/the-97-problem) — Why multi-agent systems fail to scale |
|
|
|
|
|
## Implementation |
|
|
|
|
|
- **hct-core**: [github.com/stefanwiest/hct-core](https://github.com/stefanwiest/hct-core) — Reference implementation |
|
|
- **hct-patterns**: [github.com/stefanwiest/hct-patterns](https://github.com/stefanwiest/hct-patterns) — 15+ diagnostic patterns |
|
|
- **hct-benchmarks**: [github.com/stefanwiest/hct-benchmarks](https://github.com/stefanwiest/hct-benchmarks) — Validation suite |
|
|
|
|
|
## Author |
|
|
|
|
|
**Stefan Wiest** |
|
|
- Web: [stefanwiest.de](https://stefanwiest.de) |
|
|
- GitHub: [skew202](https://github.com/skew202) |
|
|
- Writing: [SeekingSota](https://seekingsota.com) |
|
|
|
|
|
## License |
|
|
|
|
|
MIT License — See [LICENSE](LICENSE) for details. |
|
|
|
|
|
--- |
|
|
|
|
|
*Part of the [Stefan Wiest](https://github.com/stefanwiest) ecosystem: AI Research & Engineering · Multi-Agent System Coordination* |
|
|
|
