Free forever, no credit card.Get Started for Free →
← All posts
November 15, 2026 · 6 min read

Does Customer.io Remember Between Runs?

Does Customer.io Remember Between Runs? No. A Customer.io workflow holds no memory of its own between runs. When a person enters a workflow, the workflow reads that person exactly as the record stands at that moment: the attributes currently stored on the profile, the events the person has recorded, and the segments the person currently matches. Anything an earlier run worked out survives only if that run wrote it back first, as an attribute update, a new event, or a position inside a journey t

Does Customer.io Remember Between Runs?

No. A Customer.io workflow holds no memory of its own between runs. When a person enters a workflow, the workflow reads that person exactly as the record stands at that moment: the attributes currently stored on the profile, the events the person has recorded, and the segments the person currently matches. Anything an earlier run worked out survives only if that run wrote it back first, as an attribute update, a new event, or a position inside a journey that a later step can check.

That distinction is easy to miss because Customer.io feels continuous. People, events, and workflows are all still there tomorrow, and a person halfway through a long journey resumes at the right step. The continuity lives in stored profiles and workflow definitions, not in a memory that one run hands to the next.

What does a Customer.io person actually hold?

Customer.io stores two different kinds of data about a person, and the split matters. Attributes are current state: the values the workspace currently believes about someone, such as a name, a plan, a signup date, or a usage total. An identify call or a workflow step sets attributes, and a later write replaces the earlier value. The profile shows the latest value, not the history of values.

Events are the opposite shape. An event records that something happened: signed up, created a project, invited a teammate, hit a usage limit. Events arrive through track calls with their own properties attached, they keep their timestamps, and they accumulate. Segments and workflow conditions can ask whether an event happened, how often, and how recently. Objects add a third record type beside the person: a company, a course, a cart, or another entity the person relates to, holding its own attributes and its own relationships to people.

All of that is stored data a step can check. A later workflow reads attribute values, event counts, and segment membership. It does not recover why a person was given an exception, which onboarding angle was already tried last month, or what was promised in a support thread last week, unless someone stored that thread as an attribute or an event first.

How do Customer.io workflows actually run?

A workflow, also called a journey, is built from a trigger, message and data blocks, delays, and branches. Triggers include an event arriving, a person matching segment conditions, a date on a profile or an object arriving, a webhook, or a change to an object or a relationship. Inside the flow, delays wait for a fixed time, until a day and time, or until a condition turns true. True/false branches, multi-split branches, and random cohort branches divide people by attributes, events, and relationships. Data blocks update the person, send a new event, call a webhook and receive data back, or query objects related to the person. Every workflow also carries a goal and exit conditions, so a person leaves when the point of the journey is reached or stops applying.

Two scoping rules define what a run can see. The first is the triggering event. When an event starts a workflow, that event and its properties are available inside the journey it started, for personalisation and conditions. A different workflow, started later by a different event, does not inherit that payload. It can check that the earlier event exists on the profile. It cannot reopen what the earlier journey did with it. The second rule is the journey itself. Journey attributes can hold temporary values while a person moves through one journey, and position inside a journey persists, so a three-week onboarding flow resumes in week two. Both end at the journey boundary. A person can be inside an onboarding journey, a trial-expiry journey, and a re-engagement journey at the same time, and each tracks its own position and ignores the others.

Workflow steps can read object and relationship data, and conditions can branch on it. Updating that data from inside a workflow takes an API call through a data block. Nothing in the flow automatically turns what a run learned into a note the next run reads.

What about profile history and the activity feed?

A person profile shows a rich record: current attributes, the events recorded, the segments matched, the objects related, and the messages sent. That record answers what the workspace currently believes about a person and what that person has done.

That record is not working memory the next run consults before acting. A workflow does not open last month's journey notes to learn why a branch underperformed, and an AI step drafting the next message does not recover the earlier draft, the rejected subject line, or the reasoning behind a split from the profile. The history is there for a person to review, and for conditions to count. The next automated run starts from the profile in front of it.

The gap shows up in routine work. A scheduled agent that reviews journey performance weekly, drafts the next test, and checks it against what people were already promised has to be handed the numbers, the angle already used, and the promise already made, every week. The context arrives by hand, or not at all.

Where this leaves agents working from Customer.io

Customer.io is often one part of a wider setup. A product sends events in, an n8n, Make, or Zapier scenario updates attributes, and planning happens in a chat tool somewhere else. Each part can read the same current attributes, events, and segments. None of them inherits the conversation behind those values. So the work turns into re-briefing: background gets pasted into a prompt again, and a question answered in another tool gets asked again.

One memory that every run can read

The fix is a memory layer that lives outside any single person or journey, so every run and every tool works from the same past. Vilix AI is built for exactly that job, for scheduled agents and automations that start blank.

Vilix AI is cloud-hosted, with zero infrastructure to run or maintain. Connect it once over MCP and the same memory is available everywhere the work happens: alongside Customer.io workflows, in n8n, Make, and Zapier scenarios, and in the AI chat tools used to plan and review the work. What gets stored is full conversation history, not just facts extracted from it, so a later run can recover the exact agreement, the reason a decision was made, and the wording that was already used, not only an attribute value or an event name on a profile.

The loop is simple. Before acting, a run retrieves the relevant history for the work it will touch. It acts with that history in hand. Then it saves what it did and what it learned back to the same memory. Customer.io stays the system of record for people, events, and sends. Vilix AI becomes the system of record for the thread behind them, and the re-briefing cycle ends.

Get Started for Free

Free forever, no credit card. The free plan is free forever, the 7-day Pro trial needs no credit card, and your memory stays under your control: export everything or delete it anytime.

Get Started for Free

Persistent memory across ChatGPT, Claude, and the AI tools you already use in Vilix AI.

Get Started for Free

Free forever, no credit card.

Keep reading
Does Lemon Squeezy Remember Between Runs? The Customer Record Holds. The Run Reasoning Does Not.

Does Lemon Squeezy Remember Between Runs? The Customer Record Holds. The Run Reasoning Does Not. Ask whether Lemon Squeezy remembers between runs and the answer splits cleanly in two. Lemon Squeezy remembers the commercial state of a customer relationship in detail, because that is its job as a merchant of record for digital products. It does not remember what an automation or an AI agent was thinking while it worked on that customer. Those are different kinds of memory, and mixing them up is w

Does FastSpring Remember Between Runs? The Account Record Holds. The Run Reasoning Does Not.

Does FastSpring Remember Between Runs? The Account Record Holds. The Run Reasoning Does Not. Ask whether FastSpring remembers between runs and the answer splits cleanly in two. FastSpring remembers the commercial state of a customer relationship in detail, because that is its job as a merchant of record. It does not remember what an automation or an AI agent was thinking while it worked on that customer. Those are different kinds of memory, and mixing them up is where scheduled billing workflow

Does Paddle Remember Between Runs? The Customer Record Holds. The Run Reasoning Does Not.

Does Paddle Remember Between Runs? The Customer Record Holds. The Run Reasoning Does Not. Ask whether Paddle remembers between runs and the answer splits cleanly in two. Paddle remembers the commercial state of a customer relationship in detail, because that is its job as a merchant of record. It does not remember what an automation or an AI agent was thinking while it worked on that customer. Those are different kinds of memory, and mixing them up is where scheduled billing workflows quietly s