Agents can retain knowledge across conversations using two complementary systems: Agent Memory (per-agent) and the Knowledge Base (workspace-wide).
Agent Memory
Each agent has its own set of memory files — persistent markdown documents that are loaded at the start of every conversation. Use memory to store:
- Workflow guidelines and preferences
- Frequently needed reference data
- Standard responses and templates
- Rules and constraints the agent should always follow
Managing Memory
1
Go to Admin → Agents and select an agent
2
Navigate to the Memories section
3
Click New Memory and write content in the markdown editor
4
Save — the memory takes effect on the next conversation
How Memory Works
- Pre-loaded at session start — all memory files are injected into the agent's system prompt so they take effect from the very first message
- Runtime access — the agent can also read, write, and edit memory files during a conversation
- Persistent — changes made during a conversation persist for future sessions
- Per-agent — each agent has its own isolated memory space
Example: Teaching an Agent a Workflow
The agent will follow this process every time it's asked about the weekly report, without needing to be reminded.
Knowledge Base
The Knowledge Base stores workspace-wide reference documents accessible to all agents that have the knowledge permission enabled.
Navigate to Admin → Knowledge to manage knowledge entries.
Use Cases
Define business terms so agents use consistent language
Describe what tables and columns mean in business context
Standard procedures any agent might need
Common questions and approved answers
Enabling Knowledge for Agents
- Go to Admin → Agents and select the agent
- Enable the Use Knowledge Base permission
- Save — the agent can then search and retrieve knowledge entries during conversations
Memory vs. Knowledge Base
💡
Use memory for things specific to one agent's role. Use the knowledge base for information any agent might need.