osFoundry / Platform
Knowledge graph
Notes, people, projects, dates — auto-extracted into a graph you can browse and query.
What it is
osFoundry's Knowledge graph turns the writing already in your workspace into a browsable, queryable map of who and what it's about. As you take notes, run chat rooms, add records, and connect data sources, osFoundry extracts the entities mentioned — people, organizations, projects, technologies, concepts, places, events, products, and dates — and the relationships between them (works_at, created_by, uses, depends_on, is_part_of, decided, and more). The result is a single graph you can open and explore instead of a pile of disconnected documents.
Inside osFoundry it renders as a dedicated Graph screen: a force-directed view where you search for an entity, click a node to open an inspector, and follow its edges outward with an adjustable traversal depth. Because a workspace with many projects would otherwise render as one hairball, you can filter to a single project — and when a relationship crosses into another project, that neighbor is shown and visibly marked rather than dropped, so you can see exactly where your work touches the rest of the workspace. Every node and edge carries the source documents that produced it, so you can trace any entity or claim back to the note, room, or record it came from, and visibility follows the same channel membership rules as the underlying documents.
The graph is more than a picture. It powers graph-based retrieval: instead of only matching keywords or embeddings, agents and chat can walk the relationships between your entities to answer questions that need connected context — how a person, a project, and a decision relate across months of notes. Like everything in osFoundry, this runs in a local-first desktop workspace on your own API keys, so the extraction and querying use models you choose, with no per-seat fees and your data able to stay on your device.
Key capabilities
- Auto-extracts entities — people, organizations, projects, technologies, concepts, places, events, dates — from your notes, rooms, knowledge base, records, and data sources as you work, with no manual tagging
- Browse the whole workspace as an interactive force-directed graph: search for any entity, click a node to inspect it, expand relationships, and control how many hops out the traversal follows
- Every node and edge keeps its source documents, so you can trace any fact back to the note, room, or record it came from instead of trusting an unsourced claim
- Filter to a single project while still seeing where it connects to the rest of the workspace — cross-project neighbors are shown and marked rather than hidden, so the graph never lies by omission
- Feeds graph-based retrieval: agents and chat can walk relationships between your entities to answer questions that flat keyword or vector search would miss, all with your own model keys
Use cases
- Researcher: pulls up every person, organization, and date connected to a project across months of notes, then traces each fact back to the source note it came from
- Founder: filters the graph to one initiative and instantly sees where it touches other projects — shared vendors, overlapping people, related decisions — that flat search would never surface
- Analyst: lets an agent walk entity relationships in graph-based retrieval to answer a connected question, using their own model keys with data kept on-device
Frequently asked questions
Where do the entities in the graph come from — do I have to tag anything?
No manual tagging. osFoundry auto-extracts entities and relationships from the content you already create — notes, chat rooms, your knowledge base, records, and connected data sources. People, organizations, projects, technologies, concepts, dates, and events become nodes as you work, and each one records which source documents it came from so you can verify it.
Can I scope the graph to one project, or is it always the whole workspace?
Both. The default view is workspace-wide, and you can filter to a single project to cut through the noise. When you do, osFoundry still shows relationships that cross into other projects and marks those neighbors as outside the project rather than hiding them — so you can see where a project connects to the rest of your work without the graph silently omitting anything.
Which AI model does the extraction use, and does my data leave my machine?
osFoundry is bring-your-own-key, so extraction and querying run on the model keys you provide — local, cloud, or self-hosted — with no reseller markup. It's a local-first desktop workspace, so your data can stay on your device, and graph visibility is gated by the same channel-membership rules as the documents underneath it.