Your n8n workflows fail quietly.
This catches them.
Pisama for n8n watches real workflow executions and detects the failures that slip through: nodes that time out, errors swallowed by continue-on-fail, loops that never end, payloads that explode. Fair-code and self-hostable. Your execution data stays on your machine.
Structural analysis of the workflow, runtime analysis of the execution
The structural lane reads the workflow graph itself. The runtime lane reads what actually happened in an execution: real timings, real errors, real payload sizes. Precision is tuned against thousands of real community workflows.
A node genuinely ran past its threshold in a real execution, likely timing out the caller.
A node threw during the run, including failures hidden by continue-on-fail settings.
Output payloads or item counts ballooned mid-run, the pattern behind memory and quota blowups.
A workflow graph cycle with no exit and no iteration bound. Intentional batch loops are recognized and left alone.
Control flow tangled enough to be a reliability risk, calibrated on thousands of real community workflows.
Runtime output shape changes between executions. Static guessing is deliberately disabled; only real signal fires.
Works with n8n Cloud and self-hosted, no workflow rewrites
Install the n8n-nodes-pisama node, or point any HTTP node or error workflow at the ingest endpoint.
Give the server your n8n API key and it polls recent executions on its own. No workflow edits at all.
docker compose up starts the detection server and this dashboard on your own machine.
From detection to a fix you can apply
The cloud tier generates a targeted fix for each detection, previews the exact change, and applies it to your live workflow with one approval. Every apply is snapshotted and reversible. Your n8n credentials never leave your network; the cloud sees only the traces you send it.