Centrally-managed Restic backups, built in plain Go.
Restic does the work — encrypted, deduplicated, snapshot-based. Mantrinol does the orchestration: a small agent on every machine, a single control plane to see them all, and a NATS spine that holds the loop together. Cross-platform from day one — no Windows-first then ports later. The agent has no excuse to drift.
One Go binary per machine. Runs as a service. Wraps the
Restic CLI, schedules jobs, persists everything to a local
SQLite. Serves a small portal on 127.0.0.1:8080.
Tenant registry, agent inventory, job history. Issues commands over a NATS spine, watches the lifecycle traffic come back, and surfaces the lot in a single portal.
Where the bytes actually live. Local disk, NAS over UNC, S3, B2, anything Restic supports. Two distinct passwords kept apart — encryption is not network credentials.
A single Go source tree cross-compiles to Windows, macOS
and Linux with CGO_ENABLED=0. Pure-Go
dependencies throughout — no platform forks, no second-
class citizens.
The backup engine is Restic — proven, audited, encrypted
by default. Mantrinol orchestrates it, never replaces it.
You can read the snapshots with stock restic
if Mantrinol ever falls over.
A local portal on every agent for hands-on configuration, and a central portal for the fleet view. Same letterpress instrument-panel design — both look like they came out of the same workshop because they did.
Commands and status flow over per-agent NATS subjects. JetStream holds messages while agents are offline. Reconnects use exponential backoff with jitter; the UI shows “offline since 4m” until the link comes back.
Every repo, job and override is persisted on the agent. When the control plane is unreachable the schedule keeps running; when it comes back, inventory reconciles. No server, no panic.
Bearer tokens on the admin API, HMAC-SHA256 over agent-to-server NATS payloads, secrets pulled from Infisical at startup. NATS JWT per-agent isolation is the next layer up.
Done The end-to-end spine is wired: enrollment, ad-hoc backup, status, persisted job runs. Bearer + HMAC auth shipped. Self-serve install scripts and the central portal go too.
In flight Phase B — owner model for repos and jobs (local vs server), local-vault encryption of sensitive columns, inventory sync, override governance.
Phase C NATS JWT per-agent credentials so multi-tenant isolation stops being honour-system. Resume in-flight jobs after agent restart.
Phase E Restore workflows in the portal, Keycloak SSO for admins, installer generation, agent auto-update with signed manifests.