Memory

Memory is a persistent personal profile the assistant carries across conversations. Instead of starting each chat from scratch, the assistant has access to facts about you — your name, preferences, timezone, projects, standing defaults — so new conversations start warm, with context already in place.

Why memory is useful

Memory eliminates repetition and makes every conversation more personal:

• No re-asking — you don't have to remind the assistant of your name, timezone, or preferred communication style on every new chat.
• Standing preferences — the assistant knows you prefer metric units, bullet points, or a particular tone.
• Project context — ongoing work, accounts you own, clients you manage — the assistant is already caught up.
• Identity — your role, team, the stack you use.

Where to find your memories

On the sidebar under Assets, select Memory. You'll see all your remembered facts in a list, with options to add, edit, and delete them.

Adding a memory

Click the text input at the top and type a durable fact about yourself (e.g., "I prefer answers in metric units" or "I manage the Acme account"). Choose a category from the dropdown — general, preference, fact, context, or project — then click Add.

The memory is saved right away and will be injected into the system prompt of every new chat you start.

Organizing memories

Memories are sorted by:

1. Pinned — important facts pinned to the top.
2. Recently updated — what you've changed most recently.

Categories

Categories are just tags to help you organize your thoughts:

• general — catch-all (the default).
• preference — how you like things done ("answer in bullet points").
• fact — fixed information ("name is Alfred").
• context — ongoing projects or roles.
• project — a specific project you're working on.

You can create custom categories too — the assistant will learn and use them.

Pinning

Click the pin icon next to a memory to pin it. Pinned memories are always injected first into the assistant's context, so they take priority when space is tight.

Editing and deleting

Click on the memory text to edit it. Hit Enter or click outside the input to save.

Click the trash icon (visible on hover on desktop, always visible on mobile) to forget a memory.

How memory works in chat

When you start a new chat, SupaNet injects all your memories into the system prompt under a "What you remember about this user" heading. The assistant treats them as already-known background — not something to re-ask.

Memory is private to you — the assistant can't see another user's memories, and your memories never appear in shared chats.

Naming and updating

You can optionally give a memory a stable key — a slug like preferred_name or tone — so the assistant can update it in place instead of creating duplicates. For example:

• First chat: "Remember: preferred name is Al."
• Later chat: "Update my preferred name to Alfred."

The assistant uses the same key to overwrite instead of piling up two memories.

What the assistant remembers

The assistant stores memories when you teach it something durable:

• Your identity: name, role, team, timezone.
• Your tools: programming languages, frameworks, the stack you prefer.
• Your preferences: communication style, output format, tone.
• Your work: projects you're on, accounts you manage, clients.

The assistant doesn't store:

• Secrets: passwords, API keys, or anything sensitive.
• One-offs: details that won't matter next time.
• Repetition: if it's already remembered, it updates instead.

Scheduled runs and automation

When you schedule an agent or task, it runs with your memories already injected, so your standing preferences and defaults apply automatically.

Webhooks and external callers don't get your memories — they only see your memories if they explicitly ask the agent to read them via tools. This protects your privacy when external sources trigger automations.