> ## Documentation Index > Fetch the complete documentation index at: https://docs.tryblend.ai/llms.txt > Use this file to discover all available pages before exploring further. # Running the full workflow > Run the saved graph from the editor as a durable, recorded workflow run. Running the full workflow executes the graph's executable nodes in dependency order. Each node starts once its upstream dependencies are ready, and progress streams back as it finishes. [User Result](/workflow/nodes/result) nodes define the returned fields but do not execute themselves. [Annotation](/workflow/nodes/annotation) nodes are also excluded because they are visual canvas helpers. A full run currently includes executable branches even when they do not lead to a User Result. The editor warns about disconnected subgraphs, but they can still run and consume credits. Remove unused branches before launching when you do not want them included. Unlike a [single-node run](/concepts/single-node-run), a full run is **recorded**. It creates a run you can find later in the run history — surfaced both in the editor and on the [Runner](/concepts/runner). A single-node run is a throwaway test; a full run is a real result you can return to. ## It runs durably A full run runs durably. Once it starts, it keeps going on its own and survives interruptions — close the tab or lose your connection, and the run carries on. Progress streams as it goes: each node lights up the moment it finishes, so a long chain fills in node by node instead of going dark until the end. ## Cost The cost of a full run is the sum of every server node that runs. Fan a chain out with a [Batch](/workflow/nodes/batch) node and the cost scales with it—ten emitted items through a chain cost roughly ten times one pass. Disconnected executable branches can also contribute to the total. Check the run count and estimate before launching a large or fanned-out graph. For file batches, **Set** counts one item per file group while **Explode** counts one item per file. ## Same engine as the Runner A full run uses the same engine as the [Runner](/concepts/runner). Running your published workflow from the Runner executes exactly the same way — only the front door differs. The editor's run button and the Runner's form are two entrances to one execution path, so what you test on the canvas is what your callers get. ## Related The published run surface — the other front door to the same engine. Where recorded runs land, and how to revisit past outputs.