Customer MemoryPremium

Give your agents a long-term memory of the people they talk to — so every conversation feels personal, not generic.

Overview

Customer Memory stores two types of information about each customer:

• Facts — Short sentences the agent learns automatically during conversations. For example: “Prefers email over phone” or “Has two kids”. The agent recalls these facts at the start of future conversations to personalize its responses.
• Tags — Labels you apply manually to group or categorize a customer profile. For example: vip, enterprise, or at-risk. Tags are never set automatically by the agent.

Both are stored per agent, so each agent maintains its own independent memory of your customers.

Enabling Customer Memory

Customer Memory is configured per agent. Open the agent settings and toggle on Customer Memory. Once enabled, the agent will begin extracting facts from new conversations automatically.

How Facts Work

At the end of every conversation turn, the agent silently reviews what was said and extracts any new long-term facts about the customer. Only facts that are genuinely useful for future personalization are saved — the agent skips general knowledge, one-off requests, and anything it already knows.

On the next conversation, those facts are injected into the agent's context before it responds. The customer doesn't see this — it simply means the agent already “knows” relevant details without being told again.

Facts can be reviewed and deleted from the Customer Memory tab in your dashboard. To remove a fact, expand a customer's profile and click the × next to any fact you want to delete.

Tags

Tags let you label customer profiles for segmentation or priority. Unlike facts, tags are always set by a person — the agent never adds or removes them on its own.

To tag a customer profile, expand their row in the Customer Memory tab. The tag editor appears inline below their facts. Click Add tag to add a new label, or click the × beside an existing tag to remove it. You can also set tags in bulk using the import feature.

Viewing Customer Profiles

Open the Customer Memory tab in the dashboard to browse all profiles the agent has built. Each row represents one customer. Click a row to expand it and see:

• All saved facts, each with a delete button.
• Current tags with an editor to add or remove them.

You can also run a natural language query over all memory using the Query Memory field. For example: “Which customers have dietary restrictions?” — the agent reads through all profiles and returns a plain-text answer.

Bulk Import

You can pre-populate customer memory by uploading a file. This is useful for migrating an existing customer database, seeding the agent before launch, or applying tags to a large set of customers at once.

To import, go to the Customer Memory tab and click Import. Upload your file and confirm. The import will show a summary of how many facts were added, how many were skipped (because they already exist), and any rows that had errors.

Importing is non-destructive — it only adds new facts and merges new tags into existing profiles. It never removes facts or tags that are already there.

Import Formats

Three file formats are supported:

Each row must have at least one fact or one tag — a row with both fields empty will be reported as an error and skipped. Facts are capped at 500 characters each.

{
  "alice@example.com": {
    "facts": ["Prefers email over phone", "Has two kids"],
    "tags": ["vip"]
  },
  "bob@example.com": {
    "tags": ["at-risk"]
  },
  "carol@example.com": {
    "facts": ["Works in finance"]
  }
}