The open DAM that fills the gap

A free, self-hostable digital asset management system that combines archival-grade metadata standards with modern local AI and a clean interface. The two things the market keeps separate, in one tool, installable in one command.

View on GitHub Read the design notes See the matrix

Keepstack in use: the library, semantic search, an asset detail view, and the dashboard

The state of DAM in 2026

A survey of the twenty systems that matter most for government, archival, and institutional buyers found a market split into two camps that never met. The full, citation-backed study is in RESEARCH.md.

Camp A

Standards-rich heritage tools

ResourceSpace, DSpace, Archivematica, Preservica, Axiell. Run national archives, libraries, and museums.

Deep standards: Dublin Core, METS, PREMIS, EAD, OAI-PMH, IIIF
Real digital preservation
But: hard installs, dated interfaces, little or no AI

Camp B

AI-rich media libraries

Immich, PhotoPrism, and the commercial Bynder and Adobe. Great tagging and UX.

Semantic search, auto-tagging, clean modern interfaces
Easy to run (the open-source ones)
But: no archival standards, or not free and self-hostable

No open-source tool holds archival standards, modern AI and UX, and free easy self-hosting at once. That empty intersection is what Keepstack is built to occupy.

How Keepstack compares

Keepstack next to seven leading systems across 21 capabilities. Keepstack was graded against its own source code, competitors against the research, then audited for overclaiming. It deliberately shows where Keepstack is only partial or still on the roadmap. Full detail in COMPARISON.md.

How it's built

One Python process, one SQLite database, one blob directory, four runtime dependencies. Full detail and more diagrams in ARCHITECTURE.md.

flowchart LR
  UI["Web client
(zero-build SPA)"] -->|HTTP| API["REST API"]
  API --> ING["Ingest"] & SRCH["Search"] & STD["Standards
DC / OAI-PMH / IIIF"] & AUTH["Auth + audit"]
  ING --> AI["AI (optional)"]
  ING --> DB[("SQLite + FTS5")] & BLOB[["Content-addressable
blob store"]]
  SRCH --> DB
  API --> BLOB

sequenceDiagram
  participant U as Client
  participant I as Ingest
  participant S as Storage
  participant DB as SQLite
  U->>I: upload file
  I->>S: hash + store (SHA-256)
  alt identical content exists
    S-->>I: reuse blob (deduped)
  else new content
    I->>DB: metadata + thumbnail + tags
    I->>DB: FTS index + embedding
  end
  I-->>U: asset ready

See it running

Real captures of the app, not mockups.

Asset detail drawer with a large preview and inline metadata editing

Semantic search results for coastal environment water

Dashboard with asset counts and a fixity check panel

Everything is one click deep, nothing is dumped at once.

Design notesMarket, gap, decisions, results ResearchTop-20 study, citation-backed ComparisonThe full capability matrix ArchitectureHow the code fits together DAM 101The concepts in plain language Decision recordsWhy the big choices were made RoadmapShipped vs planned DevlogBuild notes in order