Getting Started
Introduction
Long-term memory for AI agents — one memory across every tool you use, for you and your team
HyperMemory gives your AI agents a memory that lasts. Claude, ChatGPT, Codex, Cursor and your own agents all read from and write to the same knowledge graph, so what you tell one of them — a decision, a preference, how your project works — every one of them knows next time.
Your HyperMemory at a glance in the dashboard.
What you get
One memory, every agent
A hypergraph of people, decisions, preferences, projects and facts, shared by every connected tool and kept across conversations.
Plugins that just work
The Claude Code and Codex plugins make agents recall before every reply and save what they learn — without being asked.
Organizations
Shared team memory, searched alongside everyone's own, with roles, seats and organization-wide agent skills.
Token usage
See every agent's AI token usage — by model, tool, person and kind of work — and import exact usage from provider accounts.
Files, meetings & notes
Upload documents, bring in Google Meet and Teams meetings, and write notes — all turned into connected memory.
Agent skills
Write a skill once and deliver it to every agent you and your team use.
Get connected
Light or dark
The dashboard comes in light and dark themes — switch with the sun icon at the bottom of the sidebar.
The same dashboard in light and dark mode.
How it works
- Recall. Before replying, the agent calls
hm_recallwith your words. HyperMemory runs semantic and keyword search, re-ranks the results and returns the best memories with their connections — plus your organization’s. - Answer. The agent replies with your context: past decisions, preferences, the people and projects involved.
- Remember. The agent stores what’s new as connected nodes. HyperMemory enriches them in the background — detecting people and organizations, linking related memories, merging duplicates.