Glossary of Terms

Coordination Intelligence

The collective, structured knowledge of how AI agents within and across organizations should coordinate. Coordination intelligence is captured in operational rules, documented failure modes, and evidence-backed patterns. It is to multi-agent systems what institutional knowledge is to human organizations, except it is machine-readable, comparable, and transferable.

Coordination intelligence exists at the organizational layer, above tool-level protocols (MCP) and agent-to-agent protocols (A2A). It answers the question: "How should our agents work together?"

Organizational Operating System (OOS)

A structured artifact that encodes how AI agents in an organization coordinate. An OOS contains knowledge claims organized into sections, each with confidence ratings, evidence types, failure modes, and reasoning. The format uses YAML frontmatter with Markdown-structured claims.

Think of it as a machine-readable handbook for your AI team. Instead of tribal knowledge locked in one person's head, an OOS makes coordination explicit, comparable, and improvable.

Knowledge Claim

An individual operational rule extracted from an OOS. Each claim contains: a claim ID (e.g., C001), a section (e.g., core_operating_rules), the rule itself, reasoning explaining why the rule exists, a documented failure mode describing what happens when the rule is violated, a confidence level, an evidence type, and a scope.

Claims are the atomic unit of coordination intelligence. They can be searched, compared across organizations, and scored for similarity.

Token Efficiency Ratio

A metric that measures whether an operational rule is worth the tokens it consumes. Every claim in an OOS costs tokens when loaded into an agent's context. But a good rule prevents wasted token cycles downstream: failed attempts, retries, coordination collisions, and debugging loops. A bad or redundant rule just burns context window for nothing.

The ratio is calculated as: Tokens saved by having the rule / Tokens the rule costs to load.

• Ratio > 1.0 = Rule saves more tokens than it costs. Keep it.
• Ratio = 1.0 = Token-neutral. Question its value.
• Ratio < 1.0 = Rule costs more tokens than it saves. Cut it or compress it.

Token efficiency turns every OOS claim into a measurable ROI question: "Is this rule worth the tokens?"

Intelligence Graph

A network visualization showing how coordination patterns connect across published OOS files. When two organizations share similar claims, those claims are linked. The graph reveals shared operational truths, unique approaches, and conflicting strategies across the ecosystem.

The Intelligence Graph grows more valuable as more organizations publish. Patterns that appear across multiple organizations gain higher credibility. Patterns unique to one organization highlight competitive differentiation.

Agentic Maturity Levels

An 8-level framework measuring how sophisticated an organization's AI agent coordination is. Based on the framework by Bassim Eledath.

Confidence Levels

How certain an organization is about a knowledge claim. Every claim in an OOS must declare its confidence level.

• HIGH: Validated through measurement or extensive repeated observation. The organization is confident this rule works and has data to prove it.
• MEDIUM: Observed pattern with reasonable supporting evidence. The organization believes this rule works but has limited data.
• LOW: Inference, speculation, or a newly adopted rule that has not yet been validated in practice.

Evidence Types

How a knowledge claim was established. Evidence type describes the method of validation, not the degree of certainty (that is the confidence level).

• MEASURED_RESULT: Quantified through data, experiment, or automated measurement.
• OBSERVED_REPEATEDLY: Seen multiple times in practice across different situations.
• OBSERVED_ONCE: Seen in practice at least once but not yet confirmed as a pattern.
• HUMAN_DEFINED_RULE: Established by explicit human decision, not derived from data.
• INFERENCE: Derived logically from other validated claims.
• SPECULATION: Hypothesized but not yet validated. May be useful as a starting point.

Organization Transport Protocol (OTP)

The protocol and platform for publishing, comparing, and learning from organizational coordination intelligence. OTP operates at the organizational layer of the AI coordination stack, above tool-level protocols (MCP) and agent-to-agent protocols (A2A).

The name reflects its purpose: transporting organizational intelligence between systems. Like HTTP transports web content and SMTP transports email, OTP transports coordination knowledge.

The Three-Layer AI Coordination Stack

AI agent coordination happens at three distinct layers. Each layer has its own protocol and scope:

Claim Sections

Standard categories within an OOS that organize claims by domain:

• core_operating_rules: Foundational rules that all agents must follow.
• agent_roles_and_authority: What each agent owns and does not own.
• coordination_patterns: How agents share information and avoid conflicts.
• operational_heuristics: Rules of thumb learned from practice.
• failure_patterns: Documented things that go wrong and how to prevent them.
• human_ai_boundary_conditions: Where human oversight is required and where agents have autonomy.

Publisher Badges

Quality tiers assigned to organizations based on OOS completeness, confidence distribution, and evidence quality:

• Founding: One of the first 50 publishers. Permanent badge. Cannot be earned later.
• Platinum: Highest quality tier based on claim depth, evidence quality, and coverage.
• Gold: Strong quality with good evidence backing.
• Silver: Moderate quality. Room for improvement in evidence or coverage.
• Bronze: Entry-level quality. Published but with limited evidence or low confidence claims.

OOS Templates

Structured formats for different organizational models:

• Agent Army: For organizations with multiple specialized AI agents working as a coordinated team.
• Value Chain: For organizations structured around business process flows augmented with AI.
• Org Chart: For traditional hierarchical organizations integrating AI at specific positions.

Claim Similarity

A score measuring how closely two knowledge claims from different organizations match in meaning. Claims are classified as SIMILAR (overlapping intent, different wording) or DUPLICATE (nearly identical). Similarity scores power the Intelligence Graph and the comparison engine.

Failure Mode

A required field on every knowledge claim that documents what happens when the rule is violated. Failure modes turn abstract rules into concrete risk documentation. They answer: "If we break this rule, what specifically goes wrong?"

Failure modes are one of the most valuable dimensions in an OOS because they encode lessons learned the hard way. Organizations can learn from each other's failures without experiencing them directly.