Cognitive Model Protocol (CMP)

CMP is an experimental interoperability protocol for explicit, evidence-based, uncertainty-aware, revisable collaboration models between humans and AI systems.

Current draft: 0.2.0-draft

For agents and LLMs

• Start with AGENTS.md.
• Machine-readable discovery is available in cmp.manifest.yaml.
• The schema is in schema/cmp.schema.json.
• Examples are in examples/.
• CMP must not be treated as a fixed user profile or classification system.

Index

• Canonical definition
• Rationale
• Technical specification
• JSON Schema
• Examples
• Critique guide
• Agent entrypoint
• Machine-readable manifest
• Pattern library (shared collaboration axes)
• Validator: npx cmp-validate
• Agent instructions
• Issue templates

Agent Usage

Protocol discussions must happen via GitHub Discussions, not as committed repository files. See AGENT_INSTRUCTIONS.md.