Who it is forA builder whose workflow returns an error or stalls and who needs to know which node caused it.
The problemA failed multi-step run is hard to diagnose when you only see a final error message.
The outcomeA named failing node with its status, duration, and error, plus a root-cause explanation and a suggested fix.
What you need first
- A machine with Docker and Docker Compose installed, plus git and a terminal (the README says 1 GB RAM is enough)
- An API key for an LLM provider such as DeepSeek, or any OpenAI-compatible endpoint, added later through the admin UI
- A web browser to open http://localhost:5000
Steps
- Run the workflow that is failing and open its execution trace after the run ends
- Read the timeline to find the node with the error, its duration, and token usage
- Ask the canvas AI assistant why the run failed and request root cause and fix suggestions
- Apply the suggested change to the canvas, then run the workflow again
Prompt for your agent
Using LoomFlow at https://github.com/banmu123/LoomFlow, open the execution trace for the failed run <runId> of workflow <workflowName> and ask the canvas assistant for the root cause. Expected output: the failing node name, its error, and a proposed fix. Verify by re-running the workflow after applying the fix and confirming the trace shows all nodes completed.
Caveat. The assistant's root-cause answer depends on the model you configured, so confirm any suggested fix against the trace before applying it.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for banmu123/LoomFlow and "Find why a workflow run failed and fix it from the trace"
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer