osFoundry / Platform
Memory groups
Group memories by customer, project or topic. Surfaces share or scope them as needed.
What it is
Memory groups let you organize what osFoundry remembers into named collections — one per customer, project, or topic — instead of piling every fact into a single undifferentiated memory. Each group holds the context that belongs to a given piece of work: the client's naming conventions, a project's constraints and decisions, the details of a recurring topic. When you're working on that customer or project, the assistant draws on the matching group; when you're not, that context stays out of the way. It's a way to keep separate streams of work from bleeding into each other while still building up real, reusable memory over time.
Inside osFoundry, memory groups connect to the surfaces you already work in — chat and agents, apps, code, knowledge and data. A surface can be scoped to a single group so an agent or chat only sees the context relevant to that engagement, or a group can be shared across surfaces when the same knowledge should follow the work wherever it goes. Because osFoundry is a local-first, bring-your-own-key desktop workspace, these memory groups live with you rather than on a vendor's servers: the context you accumulate stays on your device, and there's no per-seat billing or reseller markup sitting between you and the models you choose to run.
This matters most when one workspace serves many clients or projects. Without scoping, an assistant tends to blur details — applying one client's preferences to another, or surfacing a project's private context where it doesn't belong. Memory groups give you an explicit boundary you control: keep a customer's context isolated for confidentiality, keep a project's decisions together so the assistant stays consistent, and share a topic group across everything when the knowledge is genuinely general. The result is memory that gets more useful as it grows, without becoming a single tangled context that's impossible to trust or clean up.
Key capabilities
- Group memories into named collections by customer, project, or topic so unrelated context never mixes together
- Scope a chat or agent to a single memory group so it only sees the context that belongs to that work
- Share a group across surfaces — chat, agents, apps, code — when the same knowledge should follow the work everywhere
- Keep client and project context isolated for confidentiality when one workspace serves many engagements
- Local-first and BYOK: your accumulated memory stays on your device, with no per-seat fees or reseller markup
Use cases
- Agency operator: keeps a separate memory group per client so each project's tone, naming, and decisions stay isolated and never leak between accounts
- Solo consultant: scopes an agent to one project's memory group so its answers stay consistent with that engagement's constraints instead of drifting across jobs
- Researcher: builds a topic group for a recurring subject and shares it across chat, code, and apps so the same background context follows the work everywhere
Frequently asked questions
How is a memory group different from just one big memory?
A single memory forces every project and client into the same context, so details bleed across unrelated work. Memory groups split that into named collections you scope per customer, project, or topic — so the assistant draws only on the context that belongs to the work in front of you, and shares a group across surfaces only when you choose to.
Can I keep one client's context from leaking into another's?
Yes. That's the main reason to scope. You put each client or project in its own group and scope the relevant chats or agents to it, so the assistant only sees that group's context. Because osFoundry is local-first, the memory itself stays on your device rather than on a vendor's servers.
Do memory groups work across chat, agents, apps, and code?
Yes. Groups connect to the surfaces you already use in osFoundry. A surface can be scoped to a single group, or a group can be shared across surfaces when the same knowledge should follow the work — you decide per group whether it stays isolated or travels with you.