A folder database is an ordinary folder with an additive database layer. Documents, source code, images, spreadsheets, PDFs, media, and database files remain normal files. They still open in Finder, terminals, IDEs, creative tools, and the applications that created them.

The database layer supplies what a filesystem alone cannot: durable identity, relationships, versions, policies, provenance, queries, lifecycle state, and safe collaboration. The result is a workspace that remains familiar to people while becoming legible and operable to AI agents.

Folders are excellent containers, but weak data models

A traditional folder knows paths, names, sizes, and modified dates. That is enough to store work, but not enough to explain it. A renamed file can look like a new object. A moved folder can lose its apparent relationship to a project. A new agent entering the workspace has to infer which files are authoritative, which are generated, which are sensitive, and which decisions are current.

Teams compensate with naming conventions, README files, spreadsheets, project-management systems, and tribal knowledge. Each workaround helps, but the underlying folder still has no durable model of the work it contains.

Databases add structure by taking files away

Conventional document databases and knowledge tools solve part of the problem by importing content into their own records. They make metadata, views, and queries easier, but the proprietary container becomes the source of truth. Export can be lossy, unsupported file types become attachments, and local tools no longer work directly against the canonical workspace.

A folder database reverses that relationship. The files remain authoritative. The database describes and coordinates them without pretending that every format is Markdown or converting every artifact into a row.

What the .folderbase layer adds

Folderbase uses a minimal engine-managed .folderbase directory to hold protocol data without taking ownership of user files. The layer can identify an object across renames, connect it to related work, record immutable versions, expose queryable metadata, apply policies, and preserve the provenance of human and agent changes.

One .folderbase root also defines one independent security and synchronization boundary. A personal Folderbase, a company Folderbase, and a customer Folderbase can be composed into one authorized workspace without collapsing into one giant permission domain.

  • Durable identity across rename and move operations
  • Queryable relationships and metadata
  • Immutable versions, history, and recovery
  • Explicit grants for people, devices, and agents
  • Reviewable, attributable Change Sets
  • Templates that guide structure without locking the schema

Why this matters for AI agents

Agents are trained to work with files, paths, repositories, and command-line tools. Giving them a normal workspace preserves that strength. But useful long-running agents also need stable identity, bounded authority, current context, conflict handling, and an auditable path from proposed work to accepted work.

A folder database joins those requirements. The agent receives ordinary files plus an exact, permissioned view of their structure and history. It can query before reading everything, propose coordinated changes, and leave a result that a person can review without decoding an opaque memory store.