Understanding agents

Guide series

Part 1 of 5 — Understanding agents (you are here)

Previous: — · Next: Creating agents

All guides — overview and full list

An agent in A1KnowHow is a saved assistant you configure once and use many times. You give it a name, instructions (called the System Prompt in the app), and an AI model. That is all an agent is—not a separate app, not a new account type.

To create one in the UI, see Creating agents or the feature overview AI Agents.

Where you use an agent

You pick an agent in two main places:

Place	What happens
Agent Chat	You open a document and chat in the side panel. The agent can search your workspace and help with that context.
Workflows	You add an Agent step to a workflow. Each run sends a prompt to the same agent. See Workflow authoring for how steps chain together.

The same agent can be used in chat today and in a workflow tomorrow. Memory and skills (covered in the next guides) work the same way in both places.

Our running example

Imagine you are helping someone plan a UK remortgage. They open chat and ask:

I’m 8 months from the end of my 2-year fix. Help me plan what to do next.

A general-purpose agent can give a helpful answer right now. It might explain timing windows, product transfers, and when to start shopping.

But if they come back next week and say:

Remind me what we agreed about my fix end date.

A plain agent has nothing to read. It did not forget on purpose—it was never told to save facts anywhere. Each chat starts without a shared notebook.

That gap is what the rest of this series fixes—using documents you already have in A1KnowHow.

What you will add (overview)

You do not need a new kind of agent. You extend the same agent with instructions that point at documents:

Memory — a normal document (often called MEMORY.md) where the agent reads facts before answering and adds new lines after. Same document next session; any agent with the same instructions can use it. → Agent memory
Skills — normal documents in a workspace that are short playbooks. You attach them under Skill disclosure on the agent; A1KnowHow injects matching skills into context. → Agent skills
Both together — one remortgage assistant that uses skills first, then memory. → Building your agent

Memory and skills are patterns, not agent types

There is no "Memory agent" button in the product. You create documents, then tell your agent in its instructions which documents to read and update. Workflow templates such as Agent - With Memory and Agent - With Skills + Memory are shortcuts that pre-fill those instructions.

Why documents matter

Everything in A1KnowHow already revolves around documents in workspaces. Memory and skills fit that model:

You create and edit them like any other document.
Your team can read them in the UI (useful for checking what the agent “knows”).
Permissions follow workspace access, like the rest of your knowledge base.

Next step

Continue to Creating agents to add your first agent in Settings, then Agent memory for the document it should read and update.

Documentation