Runs & debugging

Building a flow is half the job; the other half is seeing what it actually did. Every run BlockNext performs is recorded node by node, and the same view serves both live watching and after-the-fact debugging.

Where runs live

Organization → History lists every run, newest first, with two tabs:

• Runs — workflow executions, whichever way they were started: manually from the editor, by a trigger, or through the API.
• Tools — MCP tool calls made by assistants using your API keys.

Open a run to get the per-node breakdown. While a run is in progress the page updates over WebSocket — no refreshing.

Statuses

The same vocabulary applies to a run and to each node inside it:

Status	Meaning
pending	Accepted, waiting for a worker
scheduled	Queued to start at a specific time
running	Executing now
retrying	Failed and waiting to try again — see Retries
success	Finished without error
failed	Gave up after its last attempt
cancelled	Stopped before it finished
skipped	Never ran — its branch wasn’t taken, or the node is disabled

skipped is not a failure. Branching is per item: a condition sends each item down the branch its own comparison chose, and an output with no items skips whatever hangs off it.

Reading a node

Each node in a run shows its inputs, its outputs and its error, plus how long it took. That is usually enough to place a problem:

• Wrong data reaching a node — the fault is upstream, in the $reference feeding it. Check the producing node’s outputs.
• Empty output, no error — the node ran on an empty array. A node fed nothing runs zero times; look at what filtered the items out.
• Error from the provider — the message is passed through verbatim. Credential problems usually surface here as 401/403.

Retries and failure handling

Every node has a Settings tab in the editor with four controls that decide what happens when it goes wrong:

Setting	Effect
Max retries	How many extra attempts before the node is marked failed
Retry delay	How long to wait between attempts
Timeout	How long a single attempt may take
Continue on error	Let the rest of the flow proceed even though this node failed

Disabled on the same tab takes a node out of the run entirely — it reports skipped and its downstream nodes get nothing. Useful for narrowing a flow down while you debug it.

Re-running

Runs are immutable records, so there is no “edit and resume”. Fix the flow in the editor and start a new run — the failed one stays in history as evidence of what went wrong.