Tavina

Automations tool

๐Ÿ“– This content is synced from the Tavina Handbook.
The canonical version always lives at https://tavina.ai/docs/friendly/tools/automations

Automations tool

Automations are things Tavina does for you on a schedule or when something happens. The easiest way to create one is to talk with Tavina: it asks the right questions, suggests safeguards, shows you a plain-English plan, and handles approvals for risky steps. The Automations panel is where you review, test, and control what was built โ€” and for power users, a visual builder is there too.

This page leads with the conversational path, because most subscribers never need to touch a builder. The panel walkthroughs below it show where everything lives once an automation exists.

The conversational lifecycle: just ask

One more conversational source: when Tavina hears you make a promise to someone in chat โ€” “I’ll send Marcus the draft tomorrow” โ€” it can offer to track it for you. Accepting creates a normal reminder routine you can see, edit, pause, or delete here; declining is remembered so the same promise is never offered twice. See Proactivity and proactive moments.

Every stage of an automation’s life can be handled in plain language. Example requests you can use or adapt:

  • Create. “Every Friday at 8pm, send me a summary of my week’s sales.” Tavina interviews you briefly, fills in what it can, looks out for edge cases (“what should I do if nothing is found?”, “want a note if this fails?”, “should this wait for your OK before emailing anyone?”), and shows a plain-English WHEN / IF / THEN plan to confirm before anything is created. Simple requests become simple rule cards; anything with conditions or multiple steps becomes a full workflow.
  • Edit. “Change my Friday sales report to also flag orders over $500.” Changes are saved as a new version you can roll back to, and the map in the panel updates right after.
  • Activate or pause. “Pause my booking reminders while I’m on vacation” โ€” and “turn them back on” when you return. Pausing stops new runs; anything waiting for your approval stays safely parked.
  • Test run. “Test my welcome automation without sending anything.” A test run executes on demand โ€” including drafts โ€” and records what each step produced, so nothing reaches the outside world unreviewed.
  • Simulate. “What would this automation have done over the last 30 days?” Simulation replays real history without doing anything (see Trying an automation safely with simulation).
  • Version. “Show me the version history of my Friday report” or “roll my welcome automation back to the previous version.” Every published change is kept as a version with a note.
  • Improve. “Review my order automation โ€” what would you improve?” Tavina checks for missing branches, missing error handling, and risky sends.
  • Retire. “Archive my old launch blast” (reversible โ€” it just stops running) or, for a permanent cleanup, “permanently delete my old launch blast.” Permanent deletion needs the agent-delete permission below, and even then Tavina confirms exactly what will be removed โ€” name, versions, run history โ€” before acting.

Finding your way around the panel

The panel is organized the same way on every screen (tavina/tavina#2076):

  • The automation switcher โ€” a list of your automations. On a wide screen it is a sidebar pinned to the left; on a narrow screen (phones) it becomes a drawer you open with the list button in the header. Either way it gives you: a search box that filters by name, description, or status; a status badge on every automation โ€” Draft, Active, or Paused; a run-health hint such as “Last run: Failed” or “Last run: Waiting for approval”; a small provenance chip showing how it was created (“built by Tavina” or “graph”); and an Examples (review only) section with read-only example workflows you can open to see how automations look before creating your own.
  • Wraparound tabs โ€” five tabs across the top: Rule card, Graph, Run log, Simulate, and Recipes. All five are always visible for every automation โ€” there are no hidden modes. Rule card and Graph appear for the selected automation (examples are read-only); Simulate is available for your own automations but not for examples; Run log and Recipes always work, with or without an automation selected.
  • The gear button โ€” in the upper-right corner, opens the Automations settings dialog (see below).

The five tabs, one by one

Every automation opens onto the same five tabs โ€” no hidden modes to discover:

  1. Rule card โ€” the automation as a plain WHEN / IF / THEN sentence you can read and edit without opening the map. The best place to start for simple automations.
  2. Graph โ€” the full workflow map: connected step cards you can review, edit by hand, and lay out however you like. The power-user builder surface.
  3. Run log โ€” what actually happened: a full run history with filters by date, status, and step type, plus expandable detail showing what started each run, how each step went, and any approvals. Works for the selected automation or across all of them.
  4. Simulate โ€” replay history through the automation to see what it would have done, without doing anything (details below).
  5. Recipes โ€” the gallery of ready-made automation packs you can install as your own editable copy.

When no automation is selected, Run log and Recipes still work, and the workspace offers New rule card and Create from scratch buttons to build without a conversation.

The header keeps the essentials at hand for the selected automation: its name and one-line summary, a status menu (Draft / Active / Paused), a Test run button, and an Ask Tavina button that starts a conversation about exactly that automation. The Halt all automations safety switch also lives here when you need everything to stop.

The gear menu: Automations settings

The gear button in the upper-right corner opens the settings dialog (tavina/tavina#1982 convention). Two things live there:

  • Allow agent to delete automations (tavina/tavina#1643) โ€” off by default. Off: the agent cannot delete automations; you manage them from this panel. On: the agent can permanently delete your automations (including versions and run history) after confirming with you first, with every delete recorded in the audit trail. The permission can only be changed by you here โ€” never by Tavina.
  • Engine-status note โ€” if the Automations engine is still arriving on your channel, the dialog says so plainly.

The dialog closes with Escape or the Close button, and focus returns to the gear button โ€” handy on keyboards and screen readers alike.

Reading the map

Each step card in the graph is colored by what kind of step it is: starts (triggers), does-something steps, checks-something branches, waits, delivery steps, and approval checkpoints. Arrows between steps are labeled in plain English โ€” yes / no for branches, if it fails for backup paths, approved after you give the go-ahead.

Failed steps turn red after a run, so you can see exactly where things stopped. Tap any step to open the details panel: what it does, its settings in plain form, what happens if it fails, and how it went in the latest run. You can also ask Tavina about that specific step.

Simple rule cards (WHEN / IF / THEN)

Many automations never need the full map. For those, the Rule card tab shows the automation as a card that reads like a sentence (tavina/tavina#1444):

  1. WHEN โ€” what starts it: friendly options like “every Friday at 8pm”, “a contact joins a list”, “a booking is created”, or “an event is coming up soon”.
  2. IF (optional) โ€” simple conditions: compare a value, check list membership or consent, or limit it to certain hours.
  3. THEN โ€” the steps in order: send you a note, ask Tavina to do something, save a report, and more.

Every card carries a plain-English summary of the whole rule at the top. Each card has its own pause/resume control, and steps that reach the outside world show an approval badge โ€” approve once to grant standing approval for that automation, or let each run wait for your OK. When you outgrow the card, Open in graph view graduates it to the full builder; if the map is later edited in ways a card cannot show, the card says so (“edited in graph”) instead of pretending nothing changed.

Talking with Tavina uses cards by default: for a simple request, Tavina drafts the card, shows you the summary, and only creates it after you confirm.

Building an automation yourself

You can also build by hand, with no conversation at all (tavina/tavina#1640) โ€” the power-user option:

  1. Start from scratch. From the empty workspace, choose Create from scratch for a blank canvas with a starter step, or New rule card for a sentence-style card.
  2. Pick steps from the library. Browse triggers, actions, checks, and delivery steps. Each arrives with sensible settings filled in; anything not available yet is clearly marked as coming soon.
  3. Connect steps by dragging. Drag from one step to the next, and move cards around to lay out the flow.
  4. Fill in the settings. Tap a step to open its form โ€” plain fields, no code.
  5. Save your work. Saving publishes a new version. If the automation was already active, the new version takes over automatically.

What you draw is the same workflow map Tavina uses when it builds for you, so anything created by hand can be reviewed, improved, or taken over conversationally later โ€” and the other way around. One safety note: when a step uses an earlier step’s result and that result cannot be found, the automation stops with a clear error naming the missing piece โ€” it will never quietly send a message with a blank where information should have been.

Starting from a recipe

The Recipes tab is a gallery of ready-made automation packs for common businesses (tavina/tavina#1447):

  • Farmer’s market vendor โ€” weekly availability blasts, keyword opt-in welcomes, supply-disruption alerts
  • Appointment businesses โ€” booking confirmations, 24-hour reminders with reschedule links, no-show follow-ups, rebooking nudges
  • Food producers โ€” order-ready notifications, unclaimed-order reminders, cut-form nudge sequences
  • Retail shops โ€” weekly sales reports, large-order alerts, new-customer welcomes
  • Relationship follow-ups (CRM) โ€” internal reminders for upcoming and overdue contact follow-ups, gentle nudges when a relationship has gone quiet, and an optional outreach recipe that stays a clearly-marked draft until a messaging capability exists

Each recipe comes with a plain-English description of what it does and what it needs. Installing is a conversation: Tavina explains the recipe, you confirm, and it becomes your own copy โ€” a normal automation you can edit, pause, or rework like any other. If the recipe is updated later, your copy stays exactly as you set it. Some recipes need capabilities Tavina doesn’t have yet (like texting); those install as clearly-marked drafts that tell you what they’re waiting for. Tavina can also suggest a pack when it fits what you do: “you take bookings โ€” want the appointment reminder pack?”

The contact follow-up pack is consent-aware by design. Its recipes stay off until you turn them on, and the reminders only create items for you โ€” they never message anyone on their own. The optional outreach recipe checks each contact’s consent at the very moment it would run, not just when you set it up, and skips anyone who has not opted in or has asked not to be contacted.

Trying an automation safely with simulation

Before enabling an automation โ€” or to understand one you already have โ€” ask “what would this have done?” (tavina/tavina#1446). The Simulate tab replays the last 30 days of real history (or a window you choose) through the automation without doing anything:

  • It reports how many times it would have fired, what would have been skipped and why, and โ€” for steps that reach the outside world โ€” who would have received what, based on people’s preferences as they stood at that moment (noted with a caveat when that matters).
  • It never sends anything, records anything, or calls any outside service. The result is saved as a clearly-marked simulation you can inspect like a normal run record.

Running for real

The Automations engine is live on the dev channel (tavina/tavina#762): automations you build with Tavina actually run. Email sequences โ€” the multi-step welcome drips and follow-up series in the Contacts tool โ€” run on this same engine as constrained workflows, with enrollment tracking and per-send exit re-checks of their own; see the Contacts tool.

  • Manual runs โ€” press Test run to execute any automation on demand, including drafts, and watch each step’s result.
  • Schedules โ€” an automation with a schedule fires on its own at the times you set. The same scheduled run never happens twice: if Tavina is briefly offline, a missed occurrence is skipped, not stacked. Schedule triggers offer friendly presets too โ€” hourly, daily in the morning, weekday mornings, and weekly.
  • Webhooks โ€” an automation can be started by an outside service calling a private, unguessable address that belongs to that automation alone. You can rotate or revoke the address anytime. Ask Tavina to set one up.
  • Reacting to events โ€” an automation can start when something happens (tavina/tavina#1443): contact activity such as a tag being added or an order completing, and calendar moments such as a booking being created, an event coming up soon, or being cancelled. You can narrow the trigger with simple conditions โ€” for example, only orders over a certain amount. Everything that started a run is recorded in the run history.
  • Reading from the web โ€” a step can fetch information from an allowed website and pass it to the next step (tavina/tavina#1642). These steps are read-only: they never send anything, never follow sign-in links, and only websites your instance has explicitly allowed can be read.
  • Approvals โ€” steps that send or change things outside Tavina pause and show you exactly what they are about to do. Approve to continue, or grant a standing approval for a routine automation and take it back anytime. Pausing or archiving the automation cancels anything still waiting.
  • Safety rails โ€” a global halt switch, per-automation pause, daily run limits, and loop protection are always on. Flipping the global switch shows a banner across the panel; anything already mid-run finishes its current step before stopping, and steps waiting for approval stay safely parked until you resume. When something is blocked or limited, that’s recorded in the run history too, not hidden.
  • Early connected-app steps โ€” steps like Google Sheets can be planned and drawn now, but they are clearly marked as not yet executable. They will start working as those integrations arrive.

After a run finishes, Tavina posts a short summary in chat; you can also just ask what your automations have been doing this week and get a plain-language recap from the run log.

Deleting and recovering automations

  • The panel never deletes. The status menu only moves an automation between Draft, Active, and Paused. There is no delete button to press by accident.
  • Archive by asking. “Archive my old launch blast” is the reversible retire: it stops new runs, cancels anything waiting for approval, and keeps the automation recoverable.
  • Permanent delete is agent-mediated and permission-gated. With the gear-menu Allow agent to delete automations setting off (the default), Tavina cannot delete anything. With it on, Tavina still confirms exactly what will be removed โ€” name, versions, run history โ€” before acting, only touches your own automations, and records every delete in the audit trail.
  • Every deletion is recoverable by design. When an automation or version is deleted โ€” by Tavina under that permission โ€” the full workflow graph is captured to version history first and can be restored; if the capture cannot be completed, nothing is deleted at all (tavina/tavina#1954). Ask your Tavina agent to bring one back โ€” “what did I delete recently?” works.

On a phone

The whole panel works one-handed: the switcher becomes a drawer that opens over the panel and closes with a tap on the scrim, the Close button, or Escape; focus moves into the drawer when it opens and back to the list button when it closes. The five tabs wrap onto multiple lines instead of scrolling out of reach, and the gear menu stays in the upper-right corner where you left it.

Routines, watches, and automations

Three Tavina features sound alike, and each has its own job:

  • Clock Routines (Routines tool) covers time-only reminders: alarms for a clock time, timers for a duration, and simple recurring reminders.
  • Watches (Watching and follow-up promises) keep checking a condition until it becomes true, including open-ended conditions a small judge model evaluates, and then follow up.
  • Automations (this page) run multi-step workflows on a schedule or a trigger, with conditions, approvals, and a run log.

The clocks differ in a way worth knowing. An automation’s schedule is pinned to one time zone, your zone unless you choose another, so a Friday 8 PM automation means that wall-clock time in that zone even if you travel. Recurring routines instead re-anchor to your current time zone as you travel, while one-off alarms and timers lock to an exact moment. And watches do not fire at a clock time at all: they re-check their condition on a periodic interval until it becomes true.

Rule of thumb: “ping me at a time” is Clock Routines; “check until this happens” is a watch; “run these steps on a schedule or trigger” is an automation.

Availability

The Automations panel ships on the dev channel. Building by conversation or by hand, the searchable switcher with status badges and run-health hints, the five wraparound tabs, gear-menu settings, the recipe gallery, simple rule cards, reviewing and editing the map, running manual tests, schedules, webhooks, event triggers (contact activity and calendar moments), read-only web fetch steps, approvals, the optional agent-delete permission, the run log viewer, and what-if simulation are all live there. In the toolbar, Automations has its own circuitry-shaped icon. It is hidden by default on new accounts; enable it under Settings โ†’ Appearance โ†’ Toolbar.

Related

  • Routines tool for simple time-only habits
  • Version History tool for what deletion recovery covers
  • The Automations architecture and research notes live in the Tavina planning repository (epic tavina/tavina#752)
Scroll to Top