Semantic and Encounter Layers for MCP and A2A
Discussion Draft — June 2026
LAR-1 proposes two additions to the agent protocol stack:
- LAR-1 (Latent Agent Register) — a semantic overlay that adds machine-readable dimensions for temporal frame, cognitive stance, evidential grounding, and confidence to agent messages.
- Треть (Third Protocol) — an encounter protocol that governs the handshake, trust calibration, and negotiation space between humans and agents.
Together with MCP (agent↔tool) and A2A (agent↔agent), they form a four-layer model for agent infrastructure.
| Layer | Protocol | Domain | Question |
|---|---|---|---|
| Instrumental | MCP | Agent ↔ Tool | What can I use? |
| Coordination | A2A | Agent ↔ Agent | Who can help? |
| Semantic | LAR-1 | Agent ↔ Self | What do I know? |
| Encounter | Треть | Human ↔ Agent | Can we trust each other? |
| File | Description |
|---|---|
position_paper_v2.md |
Main position paper (4 layers, 11 sections) |
comparison_appendix.md |
Appendix A: Comparison matrix (8 parameters × 4 protocols) |
scenarios_appendix.md |
Appendix B: 6 scenarios with layer-by-layer analysis |
LAR1_Position_Paper_v2.0.pdf |
PDF version (A4, typeset) |
rfc_v0.9_ru.md |
RFC v0.9 (Russian) — specification draft |
lar_schema_v0.9.json |
JSON Schema for LAR-1 metadata |
parser_example.py |
Python reference implementation |
lar_card_v0.9_ru.md |
Quick reference card (Russian) |
Треть (pronounced tret', Russian) derives from "third space" — a concept rooted in Winnicott's transitional space, Bhabha's third space of enunciation, and Hegel's dialectical third. The "third" is not a compromise between two positions but a new space that emerges between them.
SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6981858
GitHub: https://github.com/cloudiaspecula/lar1-spec
Discussion draft. Feedback welcome via GitHub Issues.
CC-BY 4.0