AI GitHub
mcp-use for connecting applications to MCP tools
Use higher-level client and server tooling for MCP-enabled applications.
Key Takeaways
- The project provides SDK and development components for working with MCP capabilities.
- Choose the SDK or client surface that matches your application language.
- Use one server with a deterministic tool.
- An agent's tool catalog is a permission boundary.
How it works
The project provides SDK and development components for working with MCP capabilities. A client can discover tools and let application code or a configured agent call them.
What you need
Choose the SDK or client surface that matches your application language. Check tool availability without an agent before adding a model or application-level workflow.
What to try first
Use one server with a deterministic tool. Test capability discovery and call the tool directly before giving a model access to the same operation.
Check what happens when a server changes its list of tools mid-test. The client should have a clear way to refresh capabilities or reject an obsolete invocation.
Where to look for failures
An agent's tool catalog is a permission boundary. Check server trust and expose only the capabilities a particular task needs.
Related Franklin pages
Use one read-only server to separate connection behavior from agent choices. A deterministic tool result makes it easier to trace failures through a larger workflow.
- Build Python agents with typed dependencies, tools and model outputs. Pydantic AI for typed agent applications
- Compose agent workflows that discover tools through MCP connections. mcp-agent for workflows around MCP servers
Browse Franklin for concrete examples of tools and services exposed through MCP. Franklin’s MCP directory
Project documentation
Try it for your workflow
Explore the project’s documentation for setup and usage, or save this page to a collection for later.
Read the project documentation ↗