Storing data between runs

Each run starts with no memory of earlier runs. To carry data from one run to the next, such as which records were already synced or the ID a record has in another app, an automation writes it to a Lookup Table or a Collector bucket. Both are built-in apps and need no connection.

Choosing where to keep data

Need	Use
A value only for the rest of this run, such as a running total inside a loop	The Variables app. Values last for one run.
A value later runs look up by key, such as a mapping from a CRM ID to a billing ID	Lookup Tables
A list of items that builds up over many runs and is processed together later, such as a daily digest	Collector
Tracking which records a polling trigger has seen	Nothing. The trigger already does this.
Storing OAuth tokens	Nothing. Connections store and refresh them.

Lookup Tables

A lookup table row is addressed by three names and holds a value:

Part	Example
Key group	`crm_to_billing`
Key	`{{ trigger.body.contact_id }}`
Value name	`billing_id`
Value	`{{ steps.create_customer.output.id }}`

A key group is only a name. Use one per purpose so rows do not collide.

Actions

Action	Does	Returns in `result`
Save value	Writes one value	Empty text. A failed save fails the step.
Save multiple values, Save multiple mapped values	Writes several values in one step (up to 20 pairs for mapped values)	Empty text
Get value	Reads the values under a key group, key and value name	A list of strings. An empty list means not found.
Get multiple values	Reads several value names at once	An object of value name to list of values
Get key	Finds the key that holds a value	A list of keys
Find records	Searches by any mix of key group, key, value name, exact value, or numeric range	A list of `{keygroup, key, value_name, value}` objects, one per stored value
Find value names	Lists the value names under a key	A list of strings
Delete single value, Delete value by value name, Delete values by key, Delete values by key group	Removes rows	Empty text. Deleting nothing still succeeds.
Step from starting number (account scope or automation scope)	Keeps a counter: the first run returns the starting number, each later run adds the step value	The new counter value

Get value always returns a list, because one value name can hold several values. Read the first value with {{ steps.get_billing_id.output.result[0] }}, and test for "not found" by checking that the list is empty.

Save value's data_replacement input decides what an existing row is replaced by: the matching value name (default), the whole key, the whole key group, or nothing, which appends another value.

Example: map IDs between two apps

Ask: "when a contact is created in the CRM, create the customer in billing unless we already did, and remember the billing ID".

Get value with key group crm_to_billing, key = the CRM contact ID, value name billing_id.
Condition: len(steps.get_billing_id.output.result) == 0.
Then: create the customer in billing, then Save value with the new billing ID and expiry 180 days.
Else: use {{ steps.get_billing_id.output.result[0] }} as the existing billing ID.

Scope

Lookup Tables actions take a scope input, except Step from starting number, whose scope is in its name. A read only sees rows written with the same scope.

Scope	Rows are shared by
This automation (default)	Runs of this automation in your account
My account	Every automation in your account
Linked accounts	Your account's linked parent and its linked children. With no linked parent, the same as My account.

Linked accounts are an Enterprise feature; see Linked and associated accounts.

Expiry

Each save sets when the row is deleted, using the expiry input:

Expiry	Use
90 days (default)	Caches, duplicate markers, short-lived state
180 days	ID mappings between apps. If a mapping expires, the next run looks the record up again and saves a new one.
365 days	Yearly state
Never	Configuration that must persist indefinitely

Saving the same row again restarts its expiry. Any other expiry text fails the step instead of falling back to a default. Expired rows are no longer returned and are removed by a regular cleanup task.

Encryption

Saved values are encrypted at rest by default. Set encrypted to false on a save only when you need Find records to search that value by numeric range, which works on unencrypted number values only. Keys, key groups and value names are always stored unencrypted, so do not put secrets in them.

Seeing what an automation stored

In the dashboard gear menu, Processing > Inspect stored data lists the rows the automation saved with This automation scope, with search on each column. Processing > Dump stored data downloads them as a CSV file.

Collector buckets

A Collector bucket holds a list of items that grows across runs until something empties it. One automation collects, and another, usually on a schedule, dumps the bucket and processes everything at once.

Operation	Type	Does
Collect items into a bucket	Action	Adds an item. A list adds each element as its own item.
Dump items from bucket	Action	Returns every item as a list, oldest first, and empties the bucket
Fetch all items from bucket in order added	Action	Returns every item without emptying the bucket
All new items in bucket	Trigger	Each time it polls and the bucket has items, runs the steps once with all items joined by an optional separator, and empties the bucket
Each new item in bucket	Trigger	Each time it polls and the bucket has items, runs the steps once per item, and empties the bucket

Both triggers have Emit just the item count?, which passes only the number of items instead of the items. The bucket is still emptied.

Example: a daily digest

Ask: "collect every new feed article during the day and email me a digest at 22:00".

Collector automation: a Feed trigger, then Collect items into a bucket with bucket newsfeed_daily and item {{ trigger.titlePlaintext }}: {{ trigger.link }}.
Digest automation: scheduled 0 22 * * *, then Dump items from bucket for newsfeed_daily, then Array to single value (Transform Data) to join the list, then the email step.
Turn both automations on. A build leaves each one off.

Dump returns a list, and most email body fields expect text, so join the list before sending it.

Bucket rules

A bucket belongs to your account, not to one automation. Any automation in the account that uses the same bucket name reads and writes the same items.
Blank bucket names become default. Prefix names by purpose, such as errors_24h, when several automations collect.
Items are deleted 90 days after they were added, whether or not anything dumped them.
Use one way of emptying a bucket. A scheduled dump and a Collector trigger on the same bucket compete and split the items between them.
A dump removes the items as it reads them. If a later step in that run fails, those items are not put back.
Buckets cannot be shared across accounts.

Troubleshooting

Symptom	Likely cause	What to do
Get value never finds a row that was saved	The save and the read use different scopes, key groups or value names	Ask Claude to compare the Save and Get settings
A mapping disappeared after a few months	The row expired	Save mappings with a longer expiry, or re-create them on a miss
The digest email is empty	The collecting automation is off, the bucket names differ, or a second automation emptied the bucket first	Check both automations are on and use exactly the same bucket name
Find records by number range returns nothing	The values were saved encrypted	Save those values with Encrypted turned off