Memory
Agents that remember
Give every agent a durable memory it reads at the start of each conversation and updates as it works: persistent knowledge that survives the chat, always yours to open, edit, and search.
Memory, not just context.
Most assistants forget everything the moment a chat ends. Here, each agent keeps a durable memory that carries from one conversation to the next.
What an agent learns in one conversation is still there in the next. Memory doesn't reset when a chat ends.
Agents write durable facts as they work, then read them back the next time you talk.
Each memory opens with an Index the agent reads every turn, so it starts already knowing what it knows.
Every agent has its own memory. Open it, read it, edit it. It's yours to curate.
Memory that survives the conversation.
An agent's knowledge lives outside any single chat. Close the tab, come back a week later, start a brand-new conversation: it still remembers what matters.
Remember, then recall.
Agents write down what they learn as they work, and read it back when it's relevant, so there's no need to re-explain the same context every time.
- Prefers concise summaries over long write-ups.
- Ships in TypeScript; deploys are behind flags.
You: "Draft the release notes."
Agent: Keeps it short, in your voice, and mentions the flag, because it remembers.
Start from the map.
Every memory has an Index: a concise, always-loaded table of contents the agent reads at the start of each turn. It starts every conversation already knowing what it knows.
Each agent's private knowledge.
Memory belongs to the agent, and the agent belongs to you. Open it any time to read, edit, or search exactly what it remembers, or switch memory off for a single chat when you want a clean slate.
Give your agents a memory.
Try 200+ models free for 7 days, then pick the plan that fits. Cancel anytime.