h2ogpte-mcp-server
Enables interaction with H2OGPTe's REST API, allowing document ingestion, collection management, and chat with AI models through MCP tools and resources.
README
H2OGPTe MCP Server
H2OGPTe MCP Server is a Model Context Protocol (MCP) server that prifor the H2OGPTe project. The server is just a local proxy forwarding traffic to REST API endpoints of H2OGPTe executed as commnad.
Installation
From Pypi
pip install h2ogpte-mcp-server
From Github
# Clone git repository
git clone https://github.com/h2oai/h2ogpte-mcp-server.git
# Use the latest release version
git checkout $(git describe --tags)
make setup
make build
# The command will install h2ogpte-mcp-server to the current python environment
make install
Usage
Environment Variables
- H2OGPTE_API_KEY - (required) H2OGPTe access key. See documentation on how to get the key.
- H2OGPTE_SERVER_URL - The url of H2OGPTe server. Default value is https://h2ogpte.genai.h2o.ai.
- H2OGPTE_ALL_ENDPOINTS_AS_TOOLS - A boolean flag, specifing whether all REST API endpoints should be represented as MCP tools. If disabled, GET endpoints will be represented as resources. Default value is
true. - H2OGPTE_ENDPOINT_SET - A set of REST API endpoints that will be used for MCP tools or resources. The default value is
all_without_async_ingest. Possible values:all- All REST API endpoints on the H2OGPTe serverall_without_async_ingest- All endpoints without asynchronous ingestion endpoints. These endpoints start and returns a job. E.g.:create_ingest_upload_jobbasic- A mininal set of endpoints for chatting with collections and ingesting new documents.custom- A set of endpoints defined by the user. If chossen, theH2OGPTE_CUSTOM_ENDPOINT_SET_FILEvariable must be set.
- H2OGPTE_CUSTOM_ENDPOINT_SET_FILE - A path to file with the list of REST API endpoints. Each endpoint name must be an a separate line. The name of the endpoint is the
operationIdattribute in REST API spec file (e.g.: https://h2ogpte.genai.h2o.ai/api-spec.yaml) - H2OGPTE_CUSTOM_ENDPOINT_SPEC_FILE - A path to OpenAPI spec file in YAML format describing REST API of the H2OGPTe server. If not specified, the file is obtained from the H2OGPTe server itself. This environement variable should be used only for debugging purposes.
Example Configuration
An example MCP server configuration for MCP clients. E.g.: Cursor, Claude Desktop
{
"mcpServers": {
"h2ogpte-mcp-server": {
"command": "h2ogpte-mcp-server",
"env": {
"H2OGPTE_API_KEY": "sk-...",
"H2OGPTE_SERVER_URL": "https://h2ogpte.genai.h2o.ai",
"H2OGPTE_ALL_ENDPOINTS_AS_TOOLS": "true"
}
}
}
}
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。