Captain Remembers Your Customers. Your Scheduled Agents Still Wake Up Blank.

Teams pick Chatwoot because they want to own their support stack. It is open source, it runs on their own servers, and it talks to every channel their customers use. Then Chatwoot shipped Captain, its built-in AI agent, and support teams got something they did not have before: an AI that remembers key details about their customers from past interactions.

That memory is real. It is also smaller than it sounds, and it ends exactly where your scheduled automations begin.

What Captain's memory actually covers

Captain is a tightly integrated set of AI features, not a standalone chatbot. It has an Assistant that chats with customers, a Co-pilot that drafts replies for human agents, an FAQ engine that spots recurring questions, and a memory feature that summarizes key conversation details so future interactions carry more context.

The memory works like this: when a customer mentions something important, Captain saves key details to that contact's notes. The next time the same customer opens a conversation, the human agent (and Captain) can see what mattered before. A customer who complained about a billing issue twice does not have to explain it a third time. That is a genuine improvement over the blank-slate chat widget.

Two things are worth knowing about the boundaries of that memory.

First, it lives inside the support context. Captain's memory is built to serve conversations: per-contact notes, past thread summaries, recurring issue tracking. It does not follow your team into the rest of the stack.

Second, self-hosted teams configure their own LLM endpoint (OpenAI, or an Ollama instance on their own hardware), which is great for data control, but it does not change what gets remembered. The memory design is still per-contact and per-conversation.

Where the gap opens: your scheduled agents

Here is the part nobody puts on the feature page. Plenty of Chatwoot teams do not just answer tickets; they run scheduled AI automations alongside the inbox. A nightly agent that scans new conversations for churn signals. A morning agent that drafts follow-ups for conversations that went quiet. A weekly agent that compiles the issues Captain flagged into a product report.

Every one of those agents wakes up blank. They do not see Captain's contact notes. They do not know which customer already mentioned the outage twice this week. A scheduled follow-up agent re-reads every thread from scratch each run, re-summarizes conversations Captain already summarized, and re-learns context that already existed ten feet away inside the same stack. The community projects that wire Chatwoot to n8n and OpenAI prove the point: their READMEs describe bolt-on memory (workflow static data, Redis buffers with TTLs) because nothing persists for them automatically.

This is the recurring pattern across support AI: the memory was designed for the conversation, not for the work around it. Gorgias remembers the ticket. Zendesk AI remembers the record. Captain remembers the customer. None of them remembers what your 6 a.m. automation did yesterday.

What a shared memory layer changes

The fix is not another integration. It is a memory layer that sits under everything and belongs to your whole stack instead of one app's inbox.

That is the model Vilix AI is built on: one cloud-hosted memory that your agents, tools, and workflows all read and write over MCP. No infrastructure to run, no Redis buffers to babysit, no static-data JSON blobs decaying inside a workflow node. When the nightly churn-scan agent reads yesterday's thread summary, the morning follow-up agent can read the same entry. When you connect it to Chatwoot, Claude, and your scheduled n8n flows, the memory follows the work, not the window.

A few things that matter in practice:

• Full conversation history, not just facts. Facts rot. "Customer is on the Pro plan" is a fact; "the customer sounded frustrated about the invoice but relaxed after the refund" is the context that actually shapes the next run. Vilix AI stores real conversations so agents can revisit them.
• Same memory everywhere. The scheduled agent, the chat assistant, the dashboard query: all read the same memory. No per-app silos.
• Zero infrastructure. Cloud-hosted, nothing to self-host, nothing to patch. That matters especially for teams already carrying the weight of a self-hosted inbox.
• No lock-in. Export everything or delete it anytime in a portable format. The memory is yours; the service holds it, not owns it.
• Free to start. The free plan is free forever, and the Pro trial runs 7 days with no credit card.

The bottom line

Captain's memory answers a narrow question well: "what do I know about this customer from our past chats?" It is worth having. But if your automations run on schedules, memory needs to answer a wider one: "what has my whole stack learned, and what did yesterday's run already do about it?"

That second question has no answer inside any single app's inbox. Give your scheduled agents one memory they all share, and they stop waking up blank.

Get started at https://vilix.ai/?utm_source=vilix-blog&utm_medium=article&utm_campaign=chatwoot-captain-remembers-customers-scheduled-agents-blank