Skip to content

Troubleshooting

David Fieser edited this page Aug 23, 2026 · 1 revision

Troubleshooting

Known rough edges and their fixes. If yours is not here, please open an issue.

lcf-mcp is not recognized

The MCP server ships in the mcp extra. Install with the extra:

pip install "lcf-strain-life[mcp]"

If you use the uvx route instead, the command is uvx --from "lcf-strain-life[mcp]" lcf-mcp and only uv needs to be installed.

The client does not show any lcf tools

  • Restart the client after editing its MCP configuration. Most clients read the file at startup.
  • Check the JSON syntax. A trailing comma or missing brace makes the whole entry silently invalid in some clients.
  • VS Code uses the key servers in .vscode/mcp.json. The mcpServers key from other clients does not work there.
  • Run the launch command yourself in a terminal. If uvx --from "lcf-strain-life[mcp]" lcf-mcp prints an error, the client sees the same failure without showing it.

Python version errors on install

The package requires Python 3.11 or newer. Check with python --version and point your environment at a newer interpreter if needed.

Windows SmartScreen blocks the exe

The release exe is currently unsigned, so SmartScreen warns on first run. Choose "More info", then "Run anyway". See the FAQ for why, and for the pip alternative.

The exe takes a long time to start

Expected for now. The exe unpacks itself on every launch. The pip install, pip install "lcf-strain-life[gui]" then lcf-gui, starts much faster.

The GUI command runs but no page opens

The app serves on a local port and normally opens the browser itself. If it does not, open the Local URL printed in the terminal, usually http://localhost:8501.

Where are my saved results?

In the store directory: the LCF_STORE_DIR environment variable if set, otherwise .lcfstore in the working directory of whatever launched the server or app. Deleting that directory deletes the saved results, and nothing else.

A result contains null where I expected a number

By design. The toolkit maps non-finite floats, NaN or infinity, to null at the MCP and store boundary so every result stays valid JSON. A null in a field means the quantity was not finite for that input, and the result's warnings usually say why.

Clone this wiki locally