MCP Jp

What mediiiiium/mcp-jp MCP server does

The mediiiiium/mcp-jp MCP server repository collects separate connectors for SaaS products used by Japanese small and medium-sized businesses, along with several global services. The listed connectors cover areas including attendance management, human resources, talent management, CRM, invoicing, e-commerce, electronic contracts, recruiting, business chat, internal wikis, customer support, payments, project management, time tracking, forms, and email delivery.

Examples include SmartHR, KING OF TIME, kaonavi, CYDAS, Mazrica Sales, invoicing services, Smaregi, CloudSign, HERP Hire, LINE WORKS, PAY.JP, Pipedrive, Freshdesk, Trello, Toggl Track, Harvest, Typeform, and SendGrid. The repository also identifies services that now have official MCP servers and says older custom connectors for those services are moved to an archive and are no longer maintained.

How it works

Each connector lives in its own directory and is distributed as an independent Python package. A connector contains package metadata, a README, an MCP server implementation, and shared-style HTTP handling code included within the package. The server communicates with the corresponding SaaS API, while common response formatting and error handling help return API failures as actionable messages rather than raw stack traces. Responses may also be shortened to avoid exceeding an agent's context window.

The collection is intended for Claude and other MCP-compatible clients. Agents can interact with connected cloud services through natural-language requests, subject to the tools implemented by the selected connector and the permissions granted to its credentials.

Setup and configuration

The mediiiiium/mcp-jp MCP server collection requires Python 3.10 or newer and an MCP-compatible client. Installation is performed inside the directory for the connector you want to use, with pip install -e .. There is no single installation command for the whole repository because the connectors are independent packages.

Client configuration is connector-specific. The README provides a Claude Desktop example using the smarthr-mcp command and environment variables for a SmartHR client ID, client secret, and subdomain. Other connectors may require different credentials and variable names, so their individual README files are the source for exact setup details. Credentials are passed through environment variables rather than embedded in the connector source.

Tools and capabilities

Available capabilities depend on the connector selected. The repository documents connectors for:

• Japanese workforce, HR, recruiting, and talent-management systems
• Accounting-adjacent, invoicing, expense, contract, payment, and commerce services
• CRM, customer support, marketing, messaging, and email platforms
• Internal knowledge, project, time-tracking, and form-management services

The project guidance recommends roughly five tools per connector, but the actual tools and supported operations must be checked in each connector's README. The repository also welcomes requests and pull requests for additional services that have public REST APIs, test credentials, and no existing official MCP implementation.

Limitations and notes

The collection is not a universal interface for every listed product. Each connector is maintained separately, has its own authentication requirements, and may expose only a subset of the underlying service API. A SaaS account, suitable API credentials, and the necessary service permissions may therefore be required.

Before adopting a connector, check whether the target service has since released an official MCP server. The repository directs users toward official implementations where available and moves superseded custom connectors into archive/. The project is licensed under MIT.