nano SIEM
SOAR

Flows

Supported in:Self-hosted

Build a flow from actions, searches, notifications, Judge steps, approval gates, conditions, and loops; test it safely; then activate it and follow its runs.

Build a flow to sequence response steps you would otherwise do by hand. You lay out the steps in the flow editor, test the flow against a real alert without changing anything, and then activate it so its trigger starts real runs.

Create a flow

Start a flow

Go to Automation → Flows and click New flow. nano creates an Untitled flow draft and opens the flow editor. To start from a description instead, click Draft with pivt.

Choose a trigger

In the block palette on the left, pick a trigger under Triggers. Selecting a trigger replaces the current one. See Triggers.

Add steps

Drag blocks from the palette onto the flow, or click a block to add it after the selected step. Select a step to set its options in the inspector on the right. Pack actions appear in the palette only for packs that are installed, enabled, and have a connection.

Validate and save

Click Validate. nano compiles the flow against your installed packs and lists any problems. Then click Save draft.

A flow can't be saved as live while an action has no connection, or uses connection: default when more than one connection could match. Saving it as a draft still works.

Step kinds

BlockWhat it doesChanges anything?
Pack actionsRuns one action from an installed pack, such as looking up a host or suspending a user.Depends on the action: read, write, or destructive
nPL searchRuns an nPL search on your data and returns rows and a summary.No
JudgeAsks the AI model a fixed question about evidence the flow already has, and returns one of your labels. See AI Judge.No
EmailSends a plain-text email to a recipient configured in Settings → SOAR.Yes
Case stepsOpens or reuses a case, adds a note, sets severity or status, adds or removes a tag, or assigns the case. No pack or connection needed. See Case steps.Yes
Approval gatePauses the run until a person approves or rejects it. See Approvals.No
ConditionTakes the first branch whose expression is true, or the default branch.No
LoopRuns the steps inside it once for each item in a list.No
Pass-throughDoes nothing. Use it as a placeholder.No

A flow can also post to a webhook endpoint configured in Settings → SOAR. The palette doesn't offer this step; add it with pivt or the API.

Keep these rules in mind:

  • One condition per flow in the editor. The editor adds one Condition block per flow.
  • Loops run one item at a time. A loop handles at most 100 items, and a run handles at most 100 loop items in total. From v2.0.33 the editor no longer offers Parallel mode. A flow saved with it keeps running items in order, and you click Run items in order before you can save it again.
  • No parallel branches. A flow with parallel branches fails validation.
  • Depth. Steps can nest at most 4 levels deep, for example a loop inside a condition.
  • Size. A flow holds at most 64 steps.

See Limits and safety for every limit.

Use data from the trigger and earlier steps

Steps read values from two places:

  • The trigger event, under flow_input. For an alert trigger, this is flow_input.alert, with fields such as severity, rule_name, and entities.src_ip. See Trigger payloads.
  • Earlier steps, under results.<step id>. For example, results.lookup.verdict reads the verdict output of the step with ID lookup.

In the inspector, pick a field from the trigger event or an earlier step, or type a fixed value. A step can't read a step that comes after it.

In a search step, write a reference in double braces where a complete value goes. Don't put quotes around it:

src_endpoint.ip={{alert.entities.src_ip}} | stats count by dst_endpoint.ip

nano inserts the value as a literal, so an alert field can't change the shape of the query. The search runs with the permissions and source visibility of the person the run acts for.

Step options

OptionWhat it does
RetryRetries a failed action up to 5 times, with a constant or growing delay. Only action steps retry. An action that isn't marked safe to retry is never resent after an unclear outcome; a person decides instead.
TimeoutEach action step has at most 15 minutes to finish.
ConnectionWhich connection an action uses. Pack default uses the pack's default connection.
Approval policyWhether the step waits for a person. See When a step needs approval.

Through the API you can also set skip_if (skip a step when an expression is true), stop_after_if (end the flow after a step), and continue_on_error (keep going when a step fails). Approval gates can't continue on error.

Test a flow

Test a flow before you activate it. A test never changes anything in your tools.

Open the run dialog

In the flow editor, click Run test.

Pick the input

Choose Real alert and pick a recent alert, choose Recent case, or enter Sample values. Advanced JSON lets you edit the inputs directly. Testing with a real alert needs soar:manage.

Start the test

Click Test with alert for a real alert, or Start dry run for a recent case or sample values.

Read the results

The editor shows each step's result next to the step, and the header shows last test with how many steps passed. Pick an earlier test from test with to see its results again.

What runs during a test:

Test with a real alertDry run with sample values or a recent case
Search and Judge steps run for real.Search and Judge steps run for real.
Read actions run against the vendor, if every request they make only reads.No vendor is called. Every action is rendered, not sent.
Write and destructive actions, and notifications, are rendered, not sent. Later steps see a placeholder for their output.Notifications are rendered, not sent.
Approval gates and steps that would need approval are recorded as would_wait, with a preview of what the approver would see. The test doesn't pause.Approval gates and steps that need approval still wait for an approver.
A step that depends on a placeholder output is skipped, and the result names the missing input. A condition that reads an approval the test doesn't wait on takes its default branch and shows not evaluated.
Case steps record the change they would make. No case is changed.Case steps record the change they would make. No case is changed.

Activate, edit, and archive

StatusWhat it means
draftBeing built. Its trigger doesn't fire. You can still test it and run it manually.
liveIts trigger starts real runs.
archivedRead-only. Its trigger doesn't fire, and it can't run.
  • Activate. Validate the flow with no errors and save it, then click Activate. Its trigger starts firing real runs. Actions still follow their approval policy.
  • Edit a live flow. Click Save. Runs that start after you save use the new definition. Runs already in progress keep the version they started with.
  • Archive. On a live flow, click Archive. The trigger stops firing. Runs already queued continue.
  • Restore. On an archived flow, click Restore as draft.

Each save increases the flow's version number (v1, v2, and so on). If someone else saved the flow since you opened it, nano refuses your save and offers Load latest. Every run records the flow version and the exact definition it ran. nano doesn't keep earlier versions as separate copies you can restore.

Run a flow manually

Click Run in the flow editor. Pick the inputs, optionally attach the run to a case, and click Run flow. A manual run carries no alert, whatever the flow's trigger is, so steps that read flow_input.alert fail validation for a manual run.

To run without sending anything, start a dry run from Run test instead.

Follow runs

Go to Automation → Runs to see every run you can see. Filter by Attention, Changes, or Running, by source, or by case, or search by run ID, action, target, case, or actor. Test runs don't appear here; they appear in the flow editor.

Open a run to see:

  • its status and, if it failed, the failure reason;
  • each step, its inputs and outputs (with secrets redacted), its approval, and its duration;
  • what the vendor request looked like, with secrets removed.

To stop a run, open it and cancel it. Canceling needs soar:run. Pending approvals in the run are canceled, and any step that may already have reached the vendor is marked ambiguous.

See Audit and troubleshooting for run statuses and failure reasons.

What's next

On this page