An automation is a per-form workflow: a trigger, and an ordered list of steps that Formspring runs for you in the background. Use automations to send conditional emails, tag submissions, fan out to internal HTTP endpoints, classify with AI, or branch based on payload contents - without writing webhook receiver code.
Lifecycle
- Create an automation on a form, choose a trigger, add steps.
- Trigger fires (e.g. submission lands).
- Run is created and queued to run in the background.
- Steps execute in order, persisting outputs to the run record.
- Replay any run from the dashboard or via the API.
Each run stores its context and cursor position, which makes delayed steps resumable. If a step returns a delay, Formspring resumes the same run and continues from the next stored cursor position when the delay expires.
Step types
Automations support these step kinds:
condition- continue only when the submission payload matches the rule set.branch- choose one of two downstream step IDs.delay- pause execution and resume automatically.action- perform work such as sending email, calling a webhook, tagging a submission, classifying with AI, or invoking an integration driver.forkandjoin- run steps side by side (see below).foreach- repeat the steps after it once for every item in a list (see below).child_workflow- start another workflow of your workspace (see below).
Use conditions for one-way gates, branches for two-path routing, and actions for side effects. Keep destructive or externally visible actions after the conditions that prove they should run.
Parallel branches
Add Parallel branches to split a run, then end the branches with Wait for branches. The steps right after the fork are dealt out to the branches in order: with branch sizes of 2 and 1, the next two steps are branch 1 and the one after them is branch 2. Each branch starts as its own run, so the branches really run at the same time rather than one after the other. The join after the last branch continues the run when every branch has finished (or, if you choose, when any one has). If a branch fails, the join fails and the branches still running are stopped. The results of the finished branches are available to the steps after the join.
A branch can hold a fork of its own, one level deep, as long as that fork and the join that closes it both sit inside the branch (count them in the size of the branch: the inner fork, its branches and its join are all steps of the outer branch). A branch cannot hold a join that closes nothing inside it, and no step may jump into a branch from outside it. Formspring checks this when you save. The run page lists each branch with its state and links back from a branch to the run that started it. Branches do not count as extra runs against your monthly run limit.
Repeating steps for every item
Add Repeat for each item to run the steps right after it once for each entry of a list, such as the line items of an order. Set List to walk to the path of the list (for example trigger.items), and Steps to repeat to how many of the following steps belong to the loop; those cards are marked "Repeated for each item of step N". Only action steps can be repeated, and they have to stay in the same branch as the loop. Inside them {{ loop.item }}, {{ loop.item.field }} and {{ loop.index }} are available.
A loop is bounded: it walks at most Most items entries (never more than 50), a longer list fails the step instead of being cut short, and one run repeats at most 200 actions in total before it pauses with guidance. Each item has its own operation id, so a retry after a failed item never repeats the emails or requests of the items that already went out. The run page lists the repeated steps with the item each one ran for.
Starting another workflow
Add Start another workflow to run a second workflow of the same workspace from this one. Choose it from the list, decide whether to Wait for it to finish (the run pauses until the other workflow ends and fails if it does not succeed) or carry on without waiting, and optionally pass fixed values along. Workflows can nest at most three levels deep, a workflow can never start itself, directly or through others, and one run starts at most ten child runs. Formspring checks the cycle and depth when you save and shows the problem on the workflow field. The run page lists the child runs with their state and links to each; a child run links back to the run that started it. A test run describes what a child workflow would do without sending anything.
Connections instead of pasted keys
An integration, webhook or HTTP request step can use a connection from your workspace instead of a pasted key or URL. The automation keeps only the name of the connection; the key stays on the connection and is read each time the step runs. Reconnecting or rotating the key therefore needs no edit to the automation, and a saved version never contains a secret. If the connection needs to be reconnected, was disconnected or was removed, the step stops with a message saying so instead of sending, and the builder, the test run and the API show the connection's state before a live run fails.
An automation also pins what each version was published against. Next to the version it keeps a hash of the definition and a hash of the data mapping, and for every connection a step names, which credential generation and which destination mapping the connection had at that moment (a counter and a digest of field names, never a secret). Rotating or reconnecting a credential does not create a new version, because every version keeps reading the live connection; instead the version history warns on the versions published before the change ("The credential of Sales Slack changed since this version was published"). Saving the automation again publishes it against what is there now and clears the warning. The same warnings are in the API and over MCP.
Three actions have no connection to choose, on purpose: Send an email, Send a text message and Classify with AI run on Formspring's own sending address, messaging account and AI service, so your workspace holds no key for them. A save that names a connection for one of them is refused with that explanation. The version detail lists them as Formspring services and says when one is not available, for example when text messages are not switched on.
Sending a failure to another step
When a step still fails after the automation's tries, the run normally ends. Open the step, then If this step fails, and choose If this step still fails, go to to hand the run to a step further down instead. Tick the kinds of failure that should go there: temporary problems, the provider asking us to slow down, a request that was refused, a rejected sign-in, or (only when you choose it) "not sure whether it went through". The cards show the route in words ("If it fails, go to step 4" and "Receives failures from step 2") and a dashed line beside them. Everything is a normal form control, so it works with the keyboard alone; the JSON tab shows the same settings (on_error_step_id, on_error_classes) if you prefer to write them.
Formspring checks the route when you save and shows the reason on the field: the step it leads to has to exist, has to come later (which is also why routes can never loop), cannot be the step itself, cannot be one of the steps a loop repeats, and inside parallel branches a route stays in its branch or goes to the join that closes it. A step that is set to keep going when it fails never uses a route. The test run follows the route too. The Hand off, retry, then notify and branch template shows the whole pattern: a call that is retried three times, then an email to the owner, then a decision whether to reassure the visitor.
Plan limits
The Free plan also allows one active automation per form. Limits follow your plan and are shown on the automations page, in the API (/automation-limits) and over MCP:
| Free | Pro | Team | Scale | |
|---|---|---|---|---|
| Runs per month | 100 | unlimited | unlimited | unlimited |
| Steps in one run | 50 | 100 | 200 | 200 |
| Runs at the same time | 2 | 4 | 8 | 10 |
| Automation emails per hour | 20 | 50 | 100 | 100 |
The limits panel also shows your AI usage: the credits you have left (or that AI is included in your plan) and how many AI actions your automations ran today against the daily AI budget. An automation's AI step counts against that daily budget; credit packs are spent by the AI generators, not by automations.
A provider budget counts the calls that went through, not the attempts: a step that fails, is skipped or is retried because the provider was down spends nothing, and one whose outcome is unknown (a timeout after the request was sent) stays counted so it is never undercounted.
You can set a lower cap for your workspace (for example to keep an agency client's cost in check), and a daily budget per provider; a cap can never go above your plan. If you need more for one workspace, contact support. Automation runs and provider actions are metered separately from your form submission quota.
Workspace template library
The workspace automations page offers a library of ready-made workspace automations: alerts for an integration that degrades, keeps failing or is about to expire, follow-ups for new contacts and deals, notices for merged and deleted companies, and more. Preview shows the trigger, the steps in plain words and the settings the template asks for (for example who to email, which starts as your own address); Install saves it as a workspace automation, switched off, and opens it in the builder so you can read and adjust it before turning it on. A template your plan cannot run says why instead of failing on save: installing uses the same checks as saving from the builder. The same library is available over the API (/workspace/automation-templates) and to agents (list_automation_templates with target_type: "workspace", get_automation_template, install_automation_template).
Queue priority
Runs started by something a person is waiting on (a response, a booking, a payment, a click) go to the interactive queue; on the Team and Scale plans they go to a priority queue that nothing else shares. Scheduled runs, and everything a workspace starts after it has started more runs in a minute than its plan allows (20 on Free, 60 on Pro, 200 on Team, 600 on Scale), go to a bulk queue with workers of its own, so an import or a mass update never delays interactive runs. A run keeps its queue for its whole life, including delays and retries, and queue_tier is shown on the run. The per-workspace limit on runs at the same time still applies on every queue.
Operational notes
Automations are designed for work that should happen after a submission is accepted. For public ingestion safety, they run in the background instead of blocking the respondent's submit request. If a workflow needs to notify a person, update a CRM, and send a webhook, the respondent still receives the normal success response while the automation continues in the background.
Review the run history after changing a workflow. It shows which trigger created the run, which steps executed, any delay or branch output, and the error message for the first non-ignored failure.
