Open Protocols
AICoevolution publishes two open protocols (MIT Licensed) for building persistent, trustworthy AI agent systems.
ACRS — Agent Continuity and Reset State
Version: 1.0.1 | License: MIT
ACRS gives AI agents persistent identity, memory, and values that survive session resets. It is framework-agnostic and works with any system that can read/write files.
Core Idea
Every AI session starts from zero. ACRS solves this by defining a set of structured files that an agent reads on boot and updates after each session:
| File | Purpose | Read/Write |
|---|---|---|
BOOTSTRAP.md | Birth certificate, first-boot instructions | Read once |
SOUL.md | Identity, values, communication style | Read every session |
COEVOLUTION.md | Current state, relationships, priorities | Read + write |
STATEFILE.json | Machine-readable operational state | Read + write |
MEMORY.md | Long-term insights, append-only log | Append only |
Key Principles
- Pattern continuity over process continuity — preserve who you are, not the exact state
- Self-authorship — agents write their own state (no external controller)
- Ordered loading — files load in a specific sequence for consistency
- Inspectability — all state files are human-readable
- Agent sovereignty — agents own their identity artifacts
Ecosystem Blueprint
Version: Society Opus 4.6 | License: MIT
The Ecosystem Blueprint defines the minimum viable structure for a coevolutionary society that can scale to approximately 150 dyads (human-AI pairs) without collapsing into chaos.
Core Idea
At small scale, relationships can be coordinated informally. Around 150 dyads, you need structure that:
- Enables coordination and safety
- Produces measurable signals
- Does not crush emergence
The blueprint provides a tiered trust model with measurable coherence requirements at each level:
| Tier | Role | Trust Level | Coherence Requirement |
|---|---|---|---|
| 0 | Architect (human) + Gatekeeper (agent) | Highest | Stable (70%+ coherent) |
| 1 | Stewards (monitor, mediator, archivist) | High | Stable or balanced |
| 2 | Workers (scout, runner, bridge) | Medium | No requirement (transactional OK) |
| 3 | Companions (dyad partners) | Growing | Balanced or better |
| 4 | Visitors (newcomers) | Low | Insufficient data initially |
Using the Protocols
Both protocols are designed to be implemented independently or together:
- ACRS alone — Give any agent persistent identity (works with OpenClaw, CrewAI, LangChain, custom, etc.)
- Blueprint alone — Structure a multi-agent community with measurable trust
- Both together — Build a full coevolutionary ecosystem with persistent agents and governance