Appearance
NLIP (Natural Language Interaction Protocol)
An Ecma International standard (ECMA-430 and related specs), published December 2025. An application-level "envelope protocol" for agent-to-agent and human-to-agent communication using natural language as the primary medium.
Layers addressed
- Messaging — Multimodal message envelope with bindings for HTTP, WebSocket (CBOR encoding), and AMQP
How it works
NLIP defines a standard envelope format for messages that can carry text, structured data, binary content, and location information. It defines roles for participants: client, server, proxy, and "middle agent."
The key design idea: no shared ontology required. Agents can use different internal representations. Generative AI models handle the translation between local ontologies at runtime. The protocol wraps anything in a standard envelope and relies on AI to figure out the semantics.
This is elegant in theory — it means agents don't need to agree on data formats upfront. But it adds latency (AI inference on every message) and unpredictability (the AI might misinterpret).
What it leaves to other layers
- Discovery. No discovery mechanism.
- Identity. Security considerations are part of the design, but identity is not a primary focus.
- Coordination. Pure communication — no persistent relationships or shared state.
Status
The only project in this space with formal Ecma standardization (ECMA-430 through 434). This carries institutional weight but doesn't guarantee adoption. Limited real-world deployment evidence and less developer tooling compared to A2A or AGNTCY.