Troubleshooting

Problem: /lamoom run <app> produces no run.

The skill is what teaches your Claude the run line. Confirm it is installed: on Claude.ai under Settings, Capabilities, Skills; on Claude Code, as a folder in ~/.claude/skills/. If you have the connector but not the skill, use the paste-in prompt on the app page instead, which works for that one app.

Problem: Adding https://console.lamoom.com/mcp does not complete sign-in.

On Claude Code, run /mcp once after adding the server. The server registers your client for you, so there is no code to paste back. If the browser step never opens, remove the server and re-add it with claude mcp add lamoom --transport http https://console.lamoom.com/mcp.

Problem: A run stopped early and left nothing behind.

A loop that writes only at the end has this failure mode. If it is your own loop, this is the first rule in Building a Loop: write reasoning and collected data at every action, not once at the finish. If it is a published app, the run trail on the console shows the last step that completed.

Problem: The loop finalized on an output you would not have shipped.

That is a missing gate rather than a bad model. Loops in the catalog refuse to finalize until their gates hold, as Well-Architected Design does with six of them. For your own loop, add the gate at the point where the weak result passed.

Problem: The console shows nothing after a run.

Runs bill and log to the signed-in account. If you ran before signing in, the run happened locally but has no account to attach to. Sign in first, then re-run.