Memory & Knowledge
CrabShack agents don't start every conversation from scratch. Two systems work together to give your agents persistent context: Memory captures what the agent learns from conversations, and the Knowledge Base gives agents access to your documents and files.
Together, these features mean your agents remember what matters and can reference the information you care about — without you repeating yourself or copying content into every conversation.
Memory
Think of Memory as a team member who takes notes during meetings and brings the relevant ones to the next conversation. As your agents work through sessions, they automatically pick up on key details — your preferences, important decisions, lessons learned, and facts worth remembering. These memories persist across conversations and are surfaced when they're relevant.
You can also add memories manually — for example, company policies or important context that might not come up naturally in conversation.
Memories are organized into tiers based on how important and frequently used they are. The most critical memories are always available to your agents, while less-used ones are recalled only when they're relevant. Over time, stale memories are automatically archived so they don't clutter the agent's context.
Knowledge Base
The Knowledge Base is like giving your agent a filing cabinet full of reference material. Upload documents — PDFs, spreadsheets, presentations, code files, and 40+ other formats — and your agents can search and reference them during conversations.
Documents are processed, indexed, and made searchable by meaning, not just keywords. When a conversation touches on a topic covered by your documents, the relevant information is automatically pulled in.
For more complex research tasks, agents can perform deep research across multiple documents, grouping and comparing information from different sources.
Cloud Drives
If your team stores documents in OneDrive or SharePoint, you can connect those services directly. CrabShack will sync your chosen folders automatically, keeping your knowledge base up to date without manual uploads.
At a Glance
| Memory | Knowledge Base | |
|---|---|---|
| Source | Learned from conversations or added manually | Uploaded files or synced from cloud drives |
| What it stores | Preferences, decisions, facts, corrections, lessons | Full documents — PDFs, spreadsheets, presentations, and more |
| How it grows | Automatically from agent conversations | You upload files or connect cloud storage |
| Who controls it | Automatic with manual override | You decide what to upload and sync |
| Shared across agents | Universal memories are shared; agent-specific memories are not | All documents are shared across all agents |
What's in This Section
- Getting Started — set up memory and knowledge in minutes
- How Memory Works — automatic extraction, context injection, and security
- Memory Types — the seven kinds of information your agent remembers
- Memory Tiers & Lifecycle — how memories are prioritized and age over time
- Managing Memories — create, review, flag, and organize memories
- How the Knowledge Base Works — document processing, indexing, and search
- Uploading Documents — supported formats, limits, and OCR
- Cloud Drives — connect OneDrive and SharePoint
- Searching & Deep Research — how agents find and use your documents
- Configuration — all settings for memory and knowledge
- Retrieval Test — test what your agent sees for any query
- Multi-Agent Behavior — how memory and knowledge work across agents
- Troubleshooting — common issues and how to fix them