Memory and Context
A generic chatbot starts every conversation from zero. MeetLoyd agents remember — they carry forward what they've learned about your business, your people, and your decisions, so they get more useful the more you work with them.
"What do you remember about the Acme account?" or "Forget the old pricing — we changed it." Loyd can recall and manage what your agents know. Meet Loyd →
Two kinds of knowing
| What it is | Example | |
|---|---|---|
| Memory | Key facts, preferences, and learnings captured from work over time | "This customer prefers email, not calls" |
| Context graph | A connected map of the people, companies, deals, and decisions agents encounter | "Jane at Acme owns the renewal, which depends on the security review" |
Memory is what an agent remembers. The context graph is how those memories connect — so an agent understands not just isolated facts, but how they relate.
How it helps
- Continuity — pick up a thread weeks later without re-explaining everything.
- Shared understanding — knowledge captured by one agent can inform its teammates, so the whole team operates from the same picture.
- Better decisions — an agent that knows the players and the history gives advice that fits your reality.
Memory grows automatically as agents work, and you can scope it to an agent, a team, or a whole workspace.
Memory, knowledge, and your control
- Memory = what agents learn from working with you.
- Documents / knowledge = reference material you bring in (manuals, policies, specs) for agents to search and cite.
Together they make agents both experienced and well-informed. And you stay in charge: memory is governed like everything else, scoped by workspace, and you can review or erase it. Where you use customer-managed keys or your own storage, that memory lives under your control.