01 / server.list
TOOLReturn every node in the registered fleet with role, region, and uptime.
params = {}
STRATUMSYS-04 / fleet-primaryUTC --:--:--
STRATUM / build 0.4.1 / rust 1.79
STRATUM
A control plane for VDS and Minecraft server fleets, written in Rust and exposed through the Model Context Protocol. One agent per node, one stream per fleet, one conversation with your model.
4 / 6NODES ONLINE2FLAGGED142CONCURRENT PLAYERS
01
Fleet
6 nodes registered, 4 healthy, 2 flagged
0x01fra-1
mc-survival-01
Survival
CPU
34%
MEM
62%
DSK
51%
- VER
- 1.21.4
- UPTIME
- 06d 11h
- PLAYERS
- 73 / 120
- KIND
- MC-SURV
0x02fra-1
mc-creative-01
Creative
CPU
09%
MEM
21%
DSK
28%
- VER
- 1.21.4
- UPTIME
- 06d 11h
- PLAYERS
- 18 / 40
- KIND
- MC-CRE
0x03fra-2
mc-modded-01
Modded
CPU
71%
MEM
84%
DSK
44%
- VER
- 1.20.1 / Fabric 0.16
- UPTIME
- 01d 03h
- PLAYERS
- 41 / 60
- KIND
- MC-MOD
0x04nyc-1
mc-network-01
Network Hub
CPU
18%
MEM
31%
DSK
22%
- VER
- Velocity 3.3.0
- UPTIME
- 00d 19h
- PLAYERS
- 10 / 80
- KIND
- MC-PROXY
0x05fra-1
vds-fra-01
Backups + Postgres
CPU
12%
MEM
39%
DSK
64%
- VER
- Debian 12 / kernel 6.1
- UPTIME
- 12d 04h
- PLAYERS
- -
- KIND
- VDS-LX
0x06fra-2
vds-fra-02
Build CI + cache
CPU
27%
MEM
51%
DSK
89%
- VER
- Debian 12 / kernel 6.1
- UPTIME
- 12d 04h
- PLAYERS
- -
- KIND
- VDS-LX
02
Live Telemetry
Aggregate metrics across the fleet, refreshed every 15s
CPU
across fleet
29%
- peak
- mc-modded-01 / 71%
- cores
- 32 eff / 64 alloc
- sample
- 15s poll
MEMORY
used of pool
48%
- peak
- mc-modded-01 / 84%
- alloc
- 94 GB / 128 GB
- swap
- 0 B
STORAGE
used of vault
50%
- peak
- vds-fra-02 / 89%
- backup
- /srv/backups 124GB free
- scrub
- ok 4m ago
03
Log Stream
Tailing wrapped systemd journal across the primary cluster
$ tail -f /var/log/stratum
18:11:42.104INFO server.bootstrap mc-survival-01 listening on 0.0.0.0:25565 (rcon=42011)
18:11:41.877INFO server.tick mc-survival-01 tick=12492 entities=4823 chunks=812 mspt=18.4
18:11:40.012WARN watchdog.cpu mc-modded-01 sustained 71 percent over 120s window
18:11:38.501INFO player.join mc-survival-01 +18 (eu-west, lobby) total=73 of 120
18:11:37.219INFO fs.scrub vds-fra-01 /srv/backups ok free=124GB prune=0
18:11:35.984INFO mcp.invoke tool=logs.read origin=claude-3.7-sonnet args=lines=40
18:11:34.000INFO alerts.dispatch 4 channels online: slack, telegram, email, pushover
18:11:30.450DEBUGnet.dns mc-network-01 resolved 3 SRV records for backend lobby
18:11:29.812INFO server.bootstrap vds-fra-02 systemd unit stratum-agent active (pid=8721)
18:11:28.001INFO alert.send channel=slack msg="mc-modded cpu high" id=al_8c4b
18:11:27.000WARN disk.headroom vds-fra-02 free=15GB of 150GB past 80 percent threshold
18:11:24.812INFO world.snapshot mc-creative-01 /srv/backups/mc-creative-01-2026-06-24.snap (3.2GB)
18:11:22.001CRIT tls.handshake admin-mac-01 cert expired 38m ago, refusing control channel
18:11:20.500INFO mcp.invoke tool=server.snapshot origin=copilot-cli args=host=vds-fra-01
18:11:18.011INFO player.leave mc-survival-01 -7 (eu-west, lobby) total=66 of 120
18:11:15.443INFO agent.rotate vds-fra-02 pickup new scrape cycle in 27s
18:11:13.012DEBUGscrape.cpu mc-modded-01 sample=%71.4 cores=4 user=58.4 sys=12.6 iow=0.4
18:11:10.992INFO channel.heartbeat slack ok telegram ok email ok pushover ok (latency=42ms p50)
18:11:08.001INFO fleet.reconcile 6 nodes registered, 0 drift, fleet checksum=7e9a80
04
Alert Center
Open incidents, sorted by severity
- CRITICAL00:01
[vds-fra-02] TLS handshake failed from admin-mac-01. Certificate expired 38 minutes ago.
id=a-901 - WARN00:12
[mc-modded-01] CPU at 71 percent sustained over 8 minutes. Chunk entity count creeping.
id=a-900 - WARN01:04
[vds-fra-02] Disk usage crossed 80 percent, holding for 6 hours.
id=a-898 - INFO00:32
[mc-survival-01] Player spike resolved: 73 of 120 concurrent, join rate returned to baseline.
id=a-897
05
MCP Surface
Tools exposed by stratum-mcp over a stdio transport
02 / server.snapshot
TOOLPull a single point-in-time read of CPU, memory, disk, and player roster.
params = { host: string }
03 / logs.read
TOOLTail the wrapped systemd journal with level and host filters.
params = { host, lines, level?, since? }
04 / logs.search
TOOLIndexed full-text search across every log buffer on a node.
params = { query, host?, since? }
05 / alert.send
TOOLFan an alert out to slack, telegram, email, or a custom webhook.
params = { severity, source, message, channels[] }
06 / player.list
TOOLReturn the live player roster with ping, world, and join timestamp.
params = { host }
07 / world.backup
TOOLSnapshot the live world to /srv/backups and rotate the previous label.
params = { host, label? }
06
Install
Wire the agent onto a fresh Debian host in about twenty seconds
// QUICK START
Install one binary per host, register a fleet, and point any Model Context Protocol client at the stdio endpoint. No stateful daemon required: the agent serves, gathers, and exits cleanly.
- ASingle Rust static binary, no glibc coupling.
- BSystemd unit, runs as the unprivileged stratum user.
- CSpeaks MCP over stdio and HTTP, your choice per client.
$ shell
# install
$ cargo install stratum-mcp --locked
# register the fleet
$ stratum init \
--token "$STRATUM_TOKEN" \
--fleet ./fleet.toml
# launch the agent
$ stratum connect
-> fleet checksum 7e9a80
-> 6 nodes registered
-> MCP endpoint http://localhost:7280/mcp
-> ready, awaiting client
Source builds from github.com/anomalyco/stratum. BSD-3-Clause.
STRATUM v0.4.1 (RUST 1.79)END OF TRANSMISSIONSIGNAL LOCKED --:--:--