io.github.icohangar-ops/chp-mcp
Enables MCP clients to evaluate CHP Profile B capital gates, including spend/mandate decisions, human-approval workflows, canonical JSON hashing, and signed audit ledger records.
README
@cubiczan/chp
Consensus Hardening Protocol for TypeScript — Profile B capital gate, canonical JSON, and signed audit ledger.
Canonical protocol repo: icohangar-ops/consensus-hardening-protocol (spec, golden vectors, Python reference)
Install
npm (this package)
npm install @cubiczan/chp
- Registry: npmjs.com/package/@cubiczan/chp
PyPI (Python reference)
pip install consensus-hardening-protocol
- Registry: pypi.org/project/consensus-hardening-protocol
- Profile A deliberation + the normative spec. This npm package is the
installable TypeScript surface for spend/mandate gates — same channel shape as
@cubiczan/codesentinel-mcp.
How the pieces fit
CHP is the engine. MCP servers are the transport.
MCP client (Cursor / Claude / …)
│ tools/call
▼
┌───────────────────────────┐
│ MCP server (transport) │ ← agent-conductor, codesentinel-mcp, …
│ decision_gate │
│ decision_adversary │
│ evaluate_spend_gate │
└─────────────┬─────────────┘
│ depends on
▼
┌───────────────────────────┐
│ Published CHP packages │
│ PyPI: consensus-hardening-protocol (Profile A)
│ npm: @cubiczan/chp (Profile B) ← you are here
└───────────────────────────┘
This package is Profile B (evaluateGate / approveHuman). Wire it into an
MCP server as something like evaluate_spend_gate. Profile A deliberation
tools (decision_gate, decision_adversary) live on
agent-conductor, which
depends on the PyPI package behind a stdio bridge.
MCP one-liner
For Cursor / Claude without writing glue code:
npx -y @cubiczan/chp-mcp
Registry name: io.github.icohangar-ops/chp-mcp. Source: cubiczan-chp-mcp.
Conformance: Profile B 30/30 · Profile A (Python) 70/70.
Quick start
import { evaluateGate, approveHuman } from "@cubiczan/chp";
const policy = {
max_notional: 500,
daily_cap: 2500,
hitl_threshold: 250,
min_confidence: 0.55,
allowed_actions: ["LONG", "SHORT"],
};
const decision = evaluateGate(
{ action: "LONG", asset: "ETH", notional: 300, confidence: 0.9 },
policy,
);
// → { state: "HITL_REQUIRED", requires_human: true, ... }
const locked = approveHuman(
{ action: "LONG", asset: "ETH", notional: 300, confidence: 0.9 },
policy,
"cfo@example.com",
);
// → { state: "LOCKED", allowed: true, ... }
Hard rule failures (max_notional, daily_cap, min_confidence, …) are
collected before the HITL threshold test. A human may cross the threshold;
they may not approve a policy violation (spec §6.3 / §6.5).
Cross-language hashes
CHP content hashes are SHA-256 over canonical JSON
(spec §3.1). Python emits whole floats as 100.0; JavaScript's JSON.stringify
emits 100. This package formats Profile B hashed field sets the Python way so
audit digests match the golden vectors and the Python reference.
Conformance adapter
npm run build
python3 path/to/spec/conformance/run_conformance.py \
--adapter-cmd "node dist/adapter.js" \
--profile B
Profile A ops are reported unsupported (SKIP), not FAIL.
Licence
MIT.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。