About Prefect Fastmcp Demos
Prefect Fastmcp Demos is an MCP server in the Science category: fastMCP demos used by the Prefect Sales Engineering team.
Install
git clone https://github.com/PrefectHQ/prefect-fastmcp-demosThis server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.
Ask AI
Ask AI about Prefect Fastmcp Demos
Powered by Claude · Grounded in docs
Security checks
- ✓README presentThe repository ships documentation.
5 checks have not been run yet.
README
Prefect FastMCP Demos
This repository contains the source code of the Prefect Sales Engineering team's FastMCP server demos.
How to use
You can use this repository as a template for building your own FastMCP projects.
Feel free to:
- Fork it into your GitHub account
- Clone it and modify it as you like
This repository uses Just to automate tasks. All recipes are in the justfile.
Add a new FastMCP server project
To create a new FastMCP server, run:
just create-server <name>
This creates a directory at servers/<name> with the appropriate template files.
Run the FastMCP server
To test your server locally, run:
just dev <name>
This starts the FastMCP server in development mode.
When you're ready to deploy, check out FastMCP.cloud, a managed platform for hosting MCP servers that helps you move quickly from development to production!
Dependency management
This repository uses uv's workspace functionality to manage dependencies across multiple FastMCP server projects. Each server can have its own dependencies specified in its pyproject.toml, while shared dependencies can be managed at the workspace level.
Additionally, constraint-dependencies are defined at the workspace level to ensure a consistent version of shared libraries across all servers.
FastMCP.cloud installs dependencies using uv pip install -r path/to/pyproject.toml. Including the constraint at the workspace level ensures correct versions are used during server builds.
How to set up a shared dependency
To add a workspace-level constraint, add the version to the [tool.uv] section in the root pyproject.toml:
[tool.uv]
constraint-dependencies = [
"fastmcp==2.3.0",
"your-dependency==x.y.z",
]
Then, in each server's pyproject.toml, you can include the dependency without specifying a version:
uv add --frozen --directory servers/<name> your-dependency
This ensures that all servers use the same version of your-dependency as specified in the workspace constraints.
README copy captured when this server was indexed; it may lag the repository. The original is authoritative.