Why doesn't the agent find a memory I did write?

Because search is by meaning. If the memory uses internal jargon and the customer asks with other words, it won't retrieve it. Write it with the words a customer would use.

What's the difference between the prompt and memories?

The prompt defines who the agent is; memories define what it knows. Prices, policies and data that changes go in memories, which you update without touching the prompt.

Can I add promotions with an expiration date?

Yes. Temporary memories have a start and end: when they expire they stop applying on their own, ideal for promotions or seasonal announcements.

Help center

How to write memories so your AI agent answers well

Memories are what your agent knows. Learn to write them so meaning-based search finds them: one idea, your customer's words and exact figures.

Published on July 18, 2026Updated on October 10, 20265 min read

What it does

Memories are your business knowledge: policies, frequently asked questions, product or service data, promotions. The agent checks them to answer with real information.

How it works

Search is by meaning, not by exact word. When the customer asks something, the engine retrieves the memories closest in meaning to the question and passes them to the agent. That's why a well-written memory uses the words a customer would use, not internal jargon.

How to write a good memory

One idea per memory. "Return policy" and "shipping times" are two memories, not one.
Self-contained. It should make sense on its own, without depending on another memory or outside context.
With your customer's words. If customers say "delivery", the memory says "delivery", not "last-mile logistics".
Specific. Exact figures, timeframes and conditions. "Free shipping from 100,000 pesos, 2 to 4 business days in major cities."

Permanent and temporary memories

Permanent ones live until they're edited or deleted. Temporary ones have a validity period (start and end) and work for promotions or seasonal announcements: when they expire, they stop applying on their own.

Where to set it up

Training menu: that's where you create, edit and delete your agent's memories.

What NOT to put in the prompt

Data that changes (prices, promos, seasonal hours). It goes in memories, which you update without touching the prompt. The prompt defines who the agent is; memories define what it knows.

Train with your own AI

Do you use ChatGPT or Claude? Connect them to your TBit agent and let them help you train it with your real data: on the platform, go to Agent Settings, the Connect your AI section, generate your token and add the connector https://api.tbit.app/mcp. Your assistant will be able to review the prompt, the memories and recent conversations, and suggest improvements.