hyperstack
由 deeqyaqub1-cmd 提供
Coding Agents & IDEs简介
The Agent Provenance Graph for AI agents — the only memory layer where agents can prove what they knew, trace why.
描述
该技能的功能
HyperStack is a typed graph memory layer for AI agents that stores facts, decisions, and their relationships with full provenance. Agents can query what breaks if a card changes, replay the exact context that existed when a decision was made, branch memory like a Git repo, and coordinate across agents via typed signals. Works via MCP, JavaScript/Python SDKs, or REST.
Stores decisions as a traversable graph rather than flat markdown logs, so agents can query impact, blockers, and provenance without parsing text or calling an LLM.
何时使用
- Tracing which services break when an auth dependency changes
- Checking blockers before deploying to production
- Replaying what an agent knew at the moment it made a specific decision
- Branching agent memory to test a new approach without affecting live state
- Coordinating multiple agents with typed signals and trust scoring
示例工作流
User asks: What breaks if we change the auth system?
- 1
Call hs_identify to register agent identity for provenance tracking
- 2
Call hs_smart_search with query 'what depends on auth?' to auto-route to impact mode
- 3
Call hs_impact with slug 'use-clerk' to find all dependent cards
- 4
Return the typed dependency list with relation labels
[auth-api] API Service via triggers, [billing-v2] Billing v2 via depends-on, [deploy-prod] Production Deploy via blocks
前置要求
- HYPERSTACK_API_KEY environment variable
- HYPERSTACK_WORKSPACE environment variable
安装
clawhub install deeqyaqub1-cmd/hyperstack运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。