the agentic startup OS v2.0.5 secure by default · sandboxed · local-first · sovereign

Linux with agents built in. On a schedule. On your machine.

vinOS is a Linux distro for founders. Autonomous agents fire 24/7 on systemd timers — your morning brief at 6 am, PR review every 30 minutes, evening recap at 6 pm. 80% runs on a local model. Free. On your metal.

the vinOS ten · live Click a theme. The whole page re-skins.

Every theme is a real palette that ships in the ISO. This chip is the same switch you'd flip on a booted vinOS box — same colors, same wallpaper.

No theme selected — using the default palette.

for

Five ways this shows up in your day.

Same ISO. Same routines. Different job to be done. Pick the angle that matches yours — each page names the pain and the fix in your language.

the money moment

Watch a routine run.

Hit play. This is a real day-brief routine executing on cosmos — scripted for the demo, but the flow is exactly what happens on your machine every morning at 06:00.

vinos-routine · day-brief
$ 
$ cost $0.000 local · $0.00 escalated
what you get

Three things that ship the day you install it.

A brief on your login screen every morning.

Reads your inbox, calendar, GitHub. Costs a penny.

day-brief · llama3.1:8b · 06:00

PR reviews while you're in a meeting.

Every 30 minutes. Local model first — Claude only for tricky diffs.

github-review · route = "auto"

Agents that travel with your code.

Drop .vinos/routines.yaml in any repo. Your team inherits the same set.

vinos-routine load . · git-native
80%
local · free
of routine work runs on a local Ollama model. No key. No round-trip.
$360/mo
saved · api bills
vs. all-premium on a typical 500-routine day. The ledger logs every call.
5 starters
shipped · disabled
day-brief · github-review · evening-shutdown · research-recap · focus-mode.
one vinos day

Six agents. All on schedule.

All quiet unless they have something worth telling you.

  1. 05:55

    hardware wake

    systemd timers arm. iwd re-associates. ollama warms llama3.1:8b in the background.

  2. 06:00

    day-brief

    top-3, signals, one reflection · opens on the login screen · llama3.1:8b · local.

  3. 09:00

    github-review

    PR queue, what needs your eyes, drafts one comment · claude · escalates when a diff exceeds 400 lines.

  4. 10:30

    focus mode

    Super+F · DND on · notifications muted · 25-min timer.

  5. 13:00

    github-review

    midday delta since 09:00 · runs local unless a PR is flagged for review.

  6. 17:00

    github-review

    end-of-day PR sweep before shutdown.

  7. 18:00

    evening-shutdown

    what shipped, what's stuck, tomorrow's top-3 · local · appended to your journal.

  8. 22:00

    research-recap

    reads ~/Reading, generates connections + spaced-repetition cards · local · claude only if 3+ new sources.

the 80/20 router

Local by default. Escalate only when needed.

Every routine picks a local Ollama model first. When the task exceeds what the local can handle — long context, hard reasoning, tool orchestration — the router hands off to a frontier API. You keep $360/mo.

local 0 premium 0 this session $0.00

Drag the slider. Each dot is a routine invocation. The router decides local or premium based on your threshold. Real math, no telemetry.

$ live ledger mock routine fires every 500 ms · 20% escalate
routed cost
$0.00
0 runs · 0 escalated
full-premium cost
$0.00
every routine on Claude
you keep
$0.00
0% saved this session
Same math as the ledger on a booted vinOS box: 20% × $0.03 avg × N runs.
.vinos/routines/day-brief.toml toml
[routine]
name        = "day-brief"
schedule    = "0 6 * * *"      # daily · 06:00 local
agent       = "brief"
route       = "auto"           # 80% local · 20% escalate

[model]
local       = "llama3.1:8b"    # ollama · default
premium     = "claude-3-5-sonnet"

[tools]
allow       = ["calendar", "inbox", "github", "weather"]

[out]
render      = "markdown"
sink        = "~/.local/state/vinos/brief.md"

[budget]
max_tokens  = 8000
max_dollars = 0.10             # per run · hard cap
laptop → cluster

Same agents. Your laptop or your kubernetes.

vinOS routines are portable. Systemd timers on your machine today; K8s CronJobs on your cluster tomorrow. Same .vinos/routines.yaml. Same 80/20 router. Same ledger. Same rules.

.vinos/routines.yaml yaml
# runs today · systemd user timer
apiVersion: vinos.computer/v1
kind: RoutineSet
metadata:
  project: startup-inc

routines:
  - name: pr-review
    schedule:
      oncalendar: "*-*-* 09,13,17:00:00"
      timezone: America/New_York
    agent:
      route: auto            # 80% local
      model: llama3.1:8b
      escalate_to: claude-sonnet-4-6
    budget:
      max_dollars_per_day: 0.50
      on_exceed: skip
k8s/pr-review-cronjob.yaml · v2.1 · vinos-cloud yaml
# same spec · exported for the cluster
apiVersion: batch/v1
kind: CronJob
metadata:
  name: pr-review
  namespace: vinos-agents
spec:
  schedule: "0 9,13,17 * * *"     # translated from oncalendar
  jobTemplate:
    spec:
      template:
        spec:
          restartPolicy: Never
          containers:
            - name: routine
              image: ghcr.io/vinpatel/vinos-cloud:2.1
              args: ["run", "pr-review"]
              envFrom:
                - secretRef: { name: vinos-api-keys }
              volumeMounts:
                - { name: state, mountPath: /var/lib/vinos }
          volumes:
            - name: state
              persistentVolumeClaim: { claimName: vinos-ledger }

vinOS Cloud (Docker + Helm) ships in v2.1 — read the roadmap →

install

Three steps.

If you have flashed a USB before, you can install this.

  1. 01

    Download the ISO.

    Grab vinos-2.0.5-x86_64.iso from the archive.org page. About 4.3 GB. Verify the sha256sum before you flash.

  2. 02

    Flash to USB.

    Etcher, Rufus, or dd. Any 8 GB+ stick works. Five minutes.

  3. 03

    Boot and install.

    Boot from the USB. When the desktop loads, open a terminal and run:

    sudo vinos-install-disk

Architecture, spec, and full docs →

Boot into your agents.

v2.0.5 is out. Flash it to a USB, boot any x86_64 machine, and in fifteen minutes your first routine is running — briefing you tomorrow morning, 80% on your own hardware.