Provides AI agents with a permanent, append-only causal history of their actions and decisions.
Sponsored
The History Graph Protocol (HGP) establishes a robust audit trail for AI agents, meticulously recording every action taken and the rationale behind it. Unlike traditional memory systems that store an agent's knowledge, HGP focuses on creating an immutable, append-only causal graph of what an agent *did*. It tracks file writes, decisions, and the evidence supporting each action, ensuring that every output is a traceable artifact. This persistence across sessions and agents allows for complete transparency, enabling the reconstruction of an agent's entire activity, understanding changes over time, and auditing decision-making processes.
主な機能
01Immutable, append-only causal graph for AI agent actions and decisions
020 GitHub stars
03Dynamic memory tiers (short-term, long-term, inactive) for operation access recency
04Four operation types: artifact, hypothesis, merge, and invalidation
05Detailed evidence trail to reconstruct decision-making processes and rationale
06Optimistic locking (leases) for managing multi-step write sequences
ユースケース
01Auditing and explaining AI agent actions and decisions for compliance or analysis
02Debugging and understanding complex AI agent behaviors across multiple sessions
03Tracing the complete provenance of AI-generated files and data artifacts
