Monitor. Investigate. Act.

Open-source fleet operations  ·  vigiles.org

The problem

• Your fleet is mixed: baremetal boxes, VMs on top, containers on top of those.
• One tool draws graphs. Another holds inventory. SSH does everything else.
• Watching and acting live in different worlds — investigation is where time dies.

What is Vigiles

• One multitenant platform that observes and controls the whole fleet.
• AI chat as the primary interface — answers grounded in live telemetry and inventory.
• Named after the Roman night watch: they watched the city and intervened.
• Open source, Apache-2.0.

Who it's for

• DevOps & sysadmins — tired of stitching tools together.
• MSPs — every client an isolated tenant.
• AI-first teams — questions first, dashboards second.

See the fleet

Every node — hosts, VMs, containers — live status, automatic host↔guest topology.

Telemetry

CPU, memory, disk, network — per host and per container, over time.

Talk to your fleet

"Which hosts have the least free disk?" — answered from live data via read tools (MCP).

Act — where allowed

A gated command catalog. Control is opt-in per host, locked until approved.

Every file, every host

FS inventory in OpenSearch: browse any host's tree, search the whole fleet in seconds.

Agents & enrollment

• Two static Go binaries: telemetry agent + control agent, one distribution.
• Token → host appears as pending → admin approves.
• Telemetry flows immediately; control stays locked until trust.
• Containers measured from the host — no agent inside.

Architecture

Multitenant by construction — every record is tenant-scoped.

Open source & roadmap

• Apache-2.0. Built in the open.
• Phase 1 — delivered: enrollment, telemetry, topology, AI chat with read tools, FS inventory.
• Phase 2: gated action commands, long-running operations (transfer, archive, fleet upgrade), async jobs + notifications.
• Contributions welcome — issues and guidelines on GitHub.

Your infrastructure never sleeps.

vigiles.org  ·  github.com/digital-cube/vigiles