Does Crisp Remember Between Runs? The Conversation Survives. The Reasoning Does Not.
Does Crisp Remember Between Runs? The Conversation Survives. The Reasoning Does Not. Short answer: Crisp keeps the conversation, not the run. A conversation keeps its messages from every channel, its private notes, its assignment, and its status. A contact keeps the person's details, their segments, and their custom data. None of them keeps what a scheduled agent worked out while reviewing that conversation, so the next run opens the same complete thread and starts its thinking from zero. Open
Does Crisp Remember Between Runs? The Conversation Survives. The Reasoning Does Not.
Short answer: Crisp keeps the conversation, not the run. A conversation keeps its messages from every channel, its private notes, its assignment, and its status. A contact keeps the person's details, their segments, and their custom data. None of them keeps what a scheduled agent worked out while reviewing that conversation, so the next run opens the same complete thread and starts its thinking from zero.
Open Crisp a month later and everything looks remembered. Old conversations are still in the inbox. The contact sidebar still shows who the person is and what they did before. Under that surface, each run starts with only the records in front of it, and none of the reasoning from the run before.
What a contact keeps
A contact in Crisp is the person behind the conversations. The sidebar shows their name, email, phone, and company when those details are available. It also shows segments and custom data, such as an account ID, when a team chose to store them.
Segments are simple labels attached to a contact or a conversation. A team can add a segment by hand, from website code, through the API, or from a workflow. A label such as lead, vip, or bug_report marks a status or a meaning, and that label stays on the record so filters, routing, and campaigns can use it later.
The sidebar can also show what a visitor did around the conversation. Last browsed pages, profile events, and device details sit in widgets next to the contact details. A returning visitor does not arrive as a stranger. A scheduled agent that reads the contact first can see the shape of the relationship before it replies.
What the contact record stays silent on is judgment. It holds no note about why a previous run decided this person needed a careful human reply instead of a fast automated one, which earlier conversation explained the current complaint, or what the next run promised to check before closing the case. The record answers who this person is and what happened with them. It says nothing about how the last run judged the situation, and that judgment is the part a scheduled review rebuilds from scratch every time.
What a conversation keeps
A conversation is where the work lives. Crisp brings live chat, email, WhatsApp, Instagram messages, Messenger, and other connected channels into one inbox, and each conversation keeps the messages exchanged on those channels in one thread.
Inside a conversation, a team keeps its working notes. Private notes add internal context that customers never see. Mentions pull a teammate into the thread. Assignment shows who owns the reply, and status and routing show where the work stands. A sub-inbox can group conversations for one team, so the queue stays clear.
That structure makes handoffs clean for people. When a conversation moves from one teammate to another, or from Crisp's AI agent Hugo to a person, the full history stays in view. The next reader sees what the customer said, what was replied, and what the team noted in between.
Read that list again and the shape is clear. A conversation keeps what was said and where the work stands. A contact keeps who the person is and how they are labeled. Neither keeps why a run escalated one conversation and closed another, which pattern across thirty conversations pointed to the same fault, or what the next review was supposed to check first. Those are decisions, and Crisp stores messages, notes, and statuses, not decisions.
Workflows remember steps, not triage
Workflows and automations hold a different kind of memory. A workflow can tag a conversation with a segment, route it to a team, or act when a conversation changes. Each step inside one workflow run can use what the steps before it collected.
That last point matters, and it has a clear boundary. Data can move forward inside a single workflow run. It does not become a shared record of reasoning that the next scheduled run can read. When a workflow finishes, its rules stay ready for the next event. The conclusions it reached about this contact, this conversation, and this morning do not stay in a form the next run starts from.
That gap grows when the review touches more than Crisp. A morning check that also reads a billing tool, a store, and a project tool has to rebuild the whole picture in every tool, because no one tool holds the reasoning that connects them. The conversation is complete inside Crisp. The thinking that linked the conversation to everything outside it leaves with the run.
What a shared memory layer adds
A shared memory layer sits outside any one tool and holds the full context of the work itself. Vilix AI is built for exactly that job. It is cloud-hosted, so there is zero infrastructure to run or maintain. The same memory is available everywhere over MCP, so a scheduled agent, a chat tool, and a coding tool all read the same shared state instead of separate copies.
Vilix AI stores full conversation history, not just extracted facts. That matters when the next run needs the reason, not only the result. The full history shows what was checked, what was ruled out, and what was promised, in the order it happened. A Crisp thread can hold where a conversation stands. Full history holds why a conversation was escalated, which earlier thread explained the complaint, and what to watch before the next reply.
So the pattern for Crisp work becomes simple. Crisp keeps doing what it does well: holding every channel in one inbox, keeping contacts, segments, and private notes in one place, and automating steps with workflows. The scheduled agent reads Vilix AI before it reviews the inbox, so it starts from the last run's conclusions instead of rebuilding them. When the run ends, it writes back what it decided and why, which conversations it cleared, which it held, and what the next run must check first. The next run, in any tool that reads the same memory, starts from that note instead of from a blank page.
The conversation record was never the problem. Crisp keeps that record with care. The missing piece was always the record of the thinking, and that piece needs a home built for it.
Free forever, no credit card. Vilix AI offers a free plan forever and a 7-day Pro trial with no credit card required, and data can be exported or deleted at any time.