Agent memory is what an AI agent remembers across messages or sessions — so it doesn't start from scratch every time you talk to it.
Agent memory is the difference between an agent that feels like a relationship and one that feels like a vending machine. Without memory, every conversation is fresh — the agent has no idea who you are or what you've talked about before. With memory, it can pick up where you left off, remember commitments, and personalise every interaction.
Memory comes in three rough flavours. Session memory holds context for one conversation — the agent remembers what you said three turns ago. Long-term memory persists across sessions — the agent remembers your name, preferences, past requests. Summarised memory compresses old context so the agent can keep useful history without the cost of carrying every word forever.
The mechanism varies. Some platforms store raw conversation logs and replay them. Others build structured memory — facts about you that get updated over time. Many do both. The right approach depends on the agent's job; a sales-development agent needs to know past conversations cold, while a quick-task assistant might do fine with session memory only.
A real-estate buyer's agent remembers from your conversation two weeks ago that you're looking to move in March, you have two kids, you need a fourth bedroom, and you can't go above £750k. When you ping it in February with "anything new?" it doesn't ask you to repeat yourself — it goes straight to listings that fit and explains why each one might or might not work for you.
Memory is what turns one-off interactions into ongoing relationships. For consumer-facing agents, that's the difference between a tool that gets used once and one that becomes a habit. For B2B agents, it's the difference between an agent that can handle a single task and one that can run a longer engagement.
For non-technical operators, the practical implication is that you should pick a platform whose memory model matches your job. If your customers come back over time, you need long-term memory. If your agent handles one-off tasks, session memory is enough — and simpler to reason about.
The risk is privacy and trust. Memory means the agent is accumulating data about each user. Where that data lives, who can see it, how a user can clear it — these are real concerns, not theoretical ones. The right platform makes the storage location, retention policies, and clear-history controls obvious.
Squidgy supports both session memory (every conversation has full context) and long-term memory (per-user facts that persist). You configure which one your agent uses when you build it. Memory is stored on Squidgy's infrastructure with clear retention controls — users can see and clear their own data, and you can configure auto-deletion windows by use case.
Memory is never shared across users by default. If you want the agent to learn from one user's interactions and apply that learning to others, that's an explicit choice you opt into.
On the platform's infrastructure, in your tenant. On Squidgy, that's our managed storage — you don't run a database. Memory is per-agent and per-user, with explicit isolation between users.
Yes — you can clear it for an individual user (e.g. on request), for a session, or in bulk by retention policy. End users can also clear their own conversation history if you expose that control to them.
By default, no — what an agent learns from one user stays with that user. You can opt into shared learning (where general patterns from many conversations improve the agent's behaviour for everyone), but it's a deliberate decision, not automatic.
No, not by default. Each user's memory is isolated. Cross-user data exposure is a serious failure mode and the platform prevents it by design.
Glossary
What is RAG (retrieval-augmented generation)?
RAG — retrieval-augmented generation — is when an AI looks up relevant info from your documents before answering, so its replies are grounded in your actual content instead of just its training data.
Glossary
What is AI agent?
An AI agent is software that takes a goal, decides what steps to take, uses tools to do them, and carries the work out with little or no human prodding between steps.
Glossary
What is Agent builder?
An agent builder is a tool for creating AI agents — defining what they do, what tools they can use, and how they decide — without writing all the code yourself.
Glossary
What is Tool calling?
Tool calling is when an AI agent decides to use a tool — like sending an email, looking up a record, or charging a card — instead of just talking about it.
No code. Hands-on onboarding from the team in your first cohort.