@general-liquidity/mcp
MCP server exposing four task-shaped tools (resolve, pay, verify, disclose) for General Liquidity, enabling agents to normalize counterparties, submit intents, verify disclosures, and produce signed disclosures.
README
@general-liquidity/mcp
A curated MCP server that projects the General Liquidity surface as four
task-shaped tools: resolve, pay, verify, and disclose. It is deliberately
not a 1:1 dump of every REST endpoint, which would overrun an agent's token
budget. The tool names are the surface verbs, and there is no settle or
grant tool: settlement stays behind the client, and mandate granting is
operator-only.
The server wraps a GeneralLiquidity client from
@general-liquidity/sdk.
The client is injected (dependency inversion): this package holds no settle
primitive and no server implementation of its own. It signs and submits intents
through the injected client; the sovereign gate decides and settles.
The four tools
resolve— normalize any counterparty reference (A2A card, signed disclosure, CAIP) into one identity with its accepted rails and trust signals.pay— submit a signed Intent to move value. The gate decides; on allow it settles on the right rail and returns a Receipt.verify— check a counterparty's signed disclosure against policy and return a Decision.disclose— produce this agent's own signed disclosure: what it is and what it is authorized to do.
Adding it to an agent host
createMcpServer returns an unconnected McpServer. You wire the client and a
transport (stdio / HTTP) at your composition root:
import { createMcpServer } from "@general-liquidity/mcp";
import { createClient } from "@general-liquidity/sdk";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
const client = createClient({ baseUrl, signer }); // any GeneralLiquidity impl
const server = createMcpServer(client, { name: "gl", version: "1.0.0" });
await server.connect(new StdioServerTransport());
Point your agent host (Claude Desktop, an IDE MCP client, or any stdio MCP host) at the process running that entry point.
What it exports
createMcpServer/McpServerOptions— build the curated MCP server over an injected client.buildTools/ToolDef/ToolResult— the tool set and its call-result shape, exported separately so tests can assert registration and delegation with a fake client and no transport.TOOL_NAMES— the four exposed tool names.
Dependencies
@general-liquidity/sdk— supplies theGeneralLiquidityclient type and the wire nouns (Intent,Disclosure,Counterparty,Receipt,Decision).@modelcontextprotocol/sdk— the MCPMcpServer.zod— tool input schemas.
Development
bun install
bunx tsc --noEmit -p tsconfig.json
bun test
bunx biome check .
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。