Skip to main content

Getting Started with Memory & Knowledge

Both Memory and the Knowledge Base are enabled by default on a fresh CrabShack installation. Your agents are already learning from conversations and ready to search any documents you provide.

What's Already On​

Out of the box:

  • Memory auto-extraction is enabled — your agents automatically learn from conversations
  • Memory injection is enabled — relevant memories are included in new conversations
  • Knowledge auto-search is enabled — if you upload documents, agents will reference them automatically

You don't need to configure anything to get started. The defaults are designed to work well for most use cases.

Guided setup

If you'd rather be walked through it, look for the guided Memory and Knowledge Base setup prompts on your dashboard after onboarding — they cover the same ground as this page in a shorter, step-by-step form.

Seeing What Your Agent Remembers​

After your agent has had a few conversations, you can see what it has learned:

  1. Open the CrabShack dashboard
  2. Navigate to Memory in the sidebar
  3. You'll see a list of everything your agent has picked up — preferences, decisions, facts, and more

Each memory shows its type, importance level, and when it was last used. If your agent hasn't had any conversations yet, this page will be empty — that's normal.

Creating a Manual Memory​

Sometimes you want your agent to know something that might not come up naturally in conversation. You can add memories manually:

  1. Navigate to Memory in the sidebar
  2. Click Add Memory
  3. Enter the content — for example, "All client reports should use the company template and be delivered as PDF"
  4. Set the importance (higher importance means the memory is more likely to be recalled)
  5. Click Save

The next time your agent handles a conversation related to reports, it will have this context available.

Uploading Your First Document​

To give your agent access to reference material:

  1. Navigate to Knowledge Base in the sidebar
  2. Drag and drop a file onto the upload area, or click to select a file
  3. Wait for the status to change from Processing to Ready
  4. Start a conversation with your agent and ask a question about the document's content

Your agent will automatically find and reference relevant sections from the document. You don't need to tell it to search — it does this on its own.

tip

Start with a document you know well so you can verify the agent is finding the right information. A company FAQ, product guide, or policy document works great for a first test.

Testing What Your Agent Sees​

The Retrieval Test tool lets you preview exactly what memories and knowledge your agent would receive for any message:

  1. Navigate to Memory or Knowledge Base in the sidebar
  2. Click the Test button (flask icon)
  3. Type a sample message and click Test
  4. Review the results — you'll see which memories and document sections would be included

This is the fastest way to verify that your agent has the right context. See Retrieval Test for a detailed guide.

What's Next​