CommonIntents-144 — the open protocol family for AI-native interaction. INTENT-7 defines what AI wants. CAPABILITY-13 defines what AI can do. Trust must be proven, not assumed.

Version: 1.0.0-draft Status: Working Group Internal Draft Date: 2026-05-22 License: Apache 2.0

Existing AI agent interaction protocols suffer from the following structural deficiencies:

• Centralized trust root: Trust anchored in a single Host node; single point of failure leads to global collapse.
• Static permission model: Permissions have no time-bound constraints; once granted, held permanently.
• Blocking human-machine collaboration: HITL is a synchronous model; the human becomes the system bottleneck.
• Abdication of security responsibility: The protocol only handles connections; security implementation is pushed downstream.
• Mandatory feature design: All features are forced upon users; no on-demand activation mechanism.

Trust must be proven, not assumed.

┌─────────────────────────────────────────┐
│              INTENT-7                         │
│  Common Intent Specification             │
│  · Pure intent semantic standard         │
│  · Transport-agnostic, crypto-agnostic   │
│  · Backend-agnostic                      │
└─────────────────────────────────────────┘
                    ▲ Semantic Binding
┌─────────────────────────────────────────┐
│              BIND-19                         │
│  INTENT-7/Transport Binding Protocol          │
│  · Transport format negotiation          │
│  · Integrity check negotiation           │
│  · Version compatibility declaration     │
└─────────────────────────────────────────┘
                    ▲ Currently Bound To
┌─────────────────────────────────────────┐
│              INTENT-7-SECURE                        │
│  Secure Intent & Control Protocol        │
│  · mTLS end-to-end encrypted transport   │
│  · Client certificate identity proof     │
└─────────────────────────────────────────┘
                    ▲ Carried By
┌─────────────────────────────────────────┐
│              CAPABILITY-13                         │
│  Capability Authentication Protocol      │
│  · Capability declaration                │
│  · Asynchronous decision queue           │
│  · Lease extension                       │
│  · Audit extension                       │
└─────────────────────────────────────────┘

The protocol defines only interaction semantic standards. Anything that does not define semantics but only mandates implementation methods is not within the protocol's scope of responsibility.

1. Minimal Core, Optional Extensions: CAPABILITY-13 Core contains only irreducible atomic functions.
2. Declarative Activation, On-Demand Execution: Advanced features are activated only when explicitly declared; not declared means zero overhead.
3. Separation of Concerns, Ecosystem Synergy: Protocol defines semantic standards; infrastructure provides implementation pipelines.
4. Zero-Dependency Ready, Progressive Enhancement: Basic compatibility requires only HTTP + JSON.
5. Maximum Embrace, Minimum Exclusion: Define only "what is correct," not "how to do it."
6. Semantically Pure, Transport-Agnostic, Backend-Agnostic: INTENT-7 is an eternal intent language, isolating all changes through BIND-19 and the static mapping layer.

All protocol specifications are published via content addressing (CID). Their authoritative versions are identified by CID, independent of any specific platform or storage service.

This constitution is maintained by the INTENT-7/CAPABILITY-13 Protocol Working Group.