History Graph Protocol FAQs

Question 1

How does HGP ensure transparency and traceability of AI decisions?

Accepted Answer

HGP achieves this through a directed acyclic causal graph linking operations and a detailed evidence trail. Any decision can cite other operations as evidence, allowing you to reconstruct the exact rationale, supporting information, and reasoning behind an agent's actions.

Question 2

What is History Graph Protocol (HGP)?

Accepted Answer

History Graph Protocol (HGP) is an MCP server that provides AI agents with a permanent, append-only causal history of their actions and decisions. It acts as an audit trail, recording 'what an agent did and why' rather than just what it knows.

Question 3

How does HGP differ from traditional AI memory systems?

Accepted Answer

Traditional AI memory systems often store what an agent *knows* (like working memory). HGP, conversely, stores what an agent *did*—every file write, decision, and the evidence supporting each action—creating a persistent, traceable audit trail across sessions.

Question 4

What types of operations can be recorded in HGP?

Accepted Answer

HGP records actions as immutable operations, categorized into four types: `artifact` (produced output like a file), `hypothesis` (a decision or plan), `merge` (combining operations), and `invalidation` (superseding a prior operation). Operations are never deleted, only added to the graph.

Question 5

What are memory tiers in HGP and how do they work?

Accepted Answer

HGP uses dynamic memory tiers (`short_term`, `long_term`, `inactive`) to manage operation access based on recency. While operations can be excluded from default queries if inactive, they are never deleted, ensuring the complete history remains accessible for auditing.

History Graph Protocol

History Graph Protocol

主要功能

使用案例

主要功能

使用案例