Concepts
Memory graphs
Separate knowledge bases for you, your projects and your organization
A memory graph is an isolated knowledge base. Nodes in one graph never connect to nodes in another, and searches don’t cross between them — except that organization members’ recall searches their organization’s graph alongside their own.
Your graphs
Every account starts with one default graph. Plans with more graphs let you keep separate knowledge bases:
| Plan | Graphs |
|---|---|
| Free, Basic | 1 |
| Pro | 4 |
| Business | 20 |
| Enterprise | Unlimited |
Manage them on the dashboard: create, rename, set the default, copy a graph’s ID, and delete.
When to use more than one
- Separate clients — keep each customer’s knowledge apart.
- Separate agents — give a support bot its own knowledge base.
- An organization’s memory — an organization’s shared memory is one of the owner’s graphs. See Organization memory.
For most people, one graph is best: HyperMemory is built to hold everything about you and connect it.
Which graph an agent uses
| Credential | Graph |
|---|---|
OAuth sign-in (plugins, connectors, hm login) | Your default graph |
| API key | The graph the key was created for |
To point an agent at a different graph, give it an API key bound to that graph — or change your default.
Organization graphs
An organization’s graph is chosen by its owner from their non-default graphs. Members read it through recall and the organization_id parameter; only admins with organization write mode write to it.
Read-only graphs after a downgrade
If you downgrade to a plan with fewer graphs, the extra graphs become read-only (frozen). API keys bound to them get 401 graph_frozen. Upgrade, or delete graphs you no longer need, to make them writable again.
Deleting a graph
Deleting a graph removes its nodes, edges, hyperedges, search indexes and files, and revokes the API keys bound to it. It can’t be undone — export first.