Enterprise

For teams and enterprises that want to own their memory system, statefulmemory self-hosts on your infra: a single daemon → per-project SQLite, git-tied anchors, and portable .mem archives. There is no vendor round-trip on the retrieval path.

Deploy

The production surface is the same statefulmemory-daemon the CLI auto-spawns. Run it in the foreground on a host, point clients at it over TCP+TLS, and attach per-project grants.

# one machine, systemd
statefulmemory daemon start --foreground

• Docker Compose, systemd, and Kubernetes manifests live in docs/enterprise/self-hosting.md (repo).
• TCP mode: statefulmemory team init-ca <dir> → mint tokens with statefulmemory team token-create --name ci --admin.
• Per-project ACLs: statefulmemory team grant --project <p> --principal <token> --role read|write gates the write handlers; admin tokens bypass.

Backup / restore

0 2 * * * cd /srv/statefulmemory && statefulmemory mem export --out backups/$(date +\%F).mem

.mem v2 archives carry the entity graph; restore with statefulmemory mem import.

Compliance evidence

Honest posture: statefulmemory is local-first. The compliance pack (docs/enterprise/ compliance.md) documents the data map, retention, deletion semantics, and the audit trail in queries.log — and explicitly does not claim certs that haven’t been earned (no SOC 2/HIPAA assertions without evidence).

Cloud

A managed StatefulMemory Cloud is on the roadmap: signup waitlist + regions TBD. Everything below the line (graph, portability, grants, UI) is self-host today. See docs/enterprise/cloud.md for the honest status.