home / mcp / mcp.science mcp server
Open source MCP server enabling AI agents to access scientific data sources and tools via a standardized protocol.
Configuration
View docs{
"mcpServers": {
"pathintegral-institute-mcp.science": {
"command": "uvx",
"args": [
"mcp-science",
"web-fetch"
]
}
}
}MCP.science provides open source MCP servers that let AI agents interact with scientific data, tools, and resources through a standardized protocol. You can start and use these servers locally to empower AI workflows with data access, computation, and targeted actions.
To use a server from MCP.science with your MCP-enabled client, start the server and connect your client to it. For example, you can run the web-fetch server and then configure your client to route queries through it. The web-fetch server fetches and processes HTML, PDF, and plain text content so your assistant can quote or summarize information from the web.
Prerequisites you need before starting: a fast package runner called uv and an MCP-enabled client application such as Claude Desktop, VSCode, Goose, or 5ire.
Install uv with this command:
curl -sSf https://astral.sh/uv/install.sh | bashOnce uv is installed, install and run the web-fetch MCP server using the standard pattern below. You only need a single command to start any server in this collection.
Start the web-fetch server using the MCP entry command:
uvx mcp-science web-fetchIntegrate the MCP server with your MCP-enabled client by pointing the client’s MCP configuration to the web-fetch server. The client will load the server as a capability it can call to fetch and process web content.
Example client configuration (illustrative, uses the standard MCP wiring pattern):
{
"mcpServers": {
"web-fetch": {
"command": "uvx",
"args": [
"mcp-science",
"web-fetch"
]
}
}
}MCPM is a convenience tool that can automate wiring servers into supported clients. It helps you manage.connections to multiple clients and servers without manual reconfiguration.
Install MCPM and set up a client, then add the web-fetch server to your configured client:
uv pip install mcpm
mcpm client ls
mcpm client set <name>
mcpm add web-fetch