agentos · command deck ONLINE · 1984 tests green
AgentOS v1 · self-hosted
the agentic operating system

Agents that do real work on your box

A login-gated platform where AI agents — Claude, on your subscription — build, fix, research and ship on your VPS. Driven from a web app and Discord, streamed live, gated by one-click approvals. Two front doors, one brain.

web https://45.136.17.227 discord two-way bot repo github · private
21
Pages
30
API groups
21
Skills
4
Models
49
Migrations
1984
Tests green
2
Front-ends
the core loop
01 · ask
Plain language
Give an agent a goal — a task, a fix, a build.
02 · act
Real tools
Shell, files, web, images — in a real workspace.
03 · stream
Live timeline
Every step replayable; nothing hidden.
04 · gate
You approve
Risky moves pause for one-click ✅ / ❌.
05 · result
Delivered
Hosted site, merged code, an answer, a file.
ways to work — the web app
DashboardMission control — live agents, approvals, inbox, spend
ChatTalk to operate the whole platform · mirrored with Discord
BoardKanban tickets — dispatch an agent, cards move themselves
LiveEverything running now, with pulsing agent org-maps
ProjectsIDE workspaces — file tree, code, runs, memory
SitesBuild & host generated apps on the box, live URLs
SchedulesCron jobs → tasks → inbox · flag one as Orchestrator
MemoryGlobal + per-project memory + the Lessons feed
SkillsThe reusable-capability library agents draw on
SessionThe run view — transcript, org-map, cost & tokens
multi-agent

Models, right-sized

Sonnet 5 default, plus Opus 4.8, Haiku 4.5, Fable 5 — per run, project, or global. Live pricing in every picker; real cost captured per run.

Subagents & teams

Agents delegate to reusable roles (researcher, coder, reviewer). A team pairs a supervisor with a roster and decomposes the work — shown as a live org-map.

The Orchestrator

A scheduled agent with full platform tools + your standing directive that manages projects and proactively messages you on Discord with updates and decisions.

Cost independence

Runs on your Claude subscription. A validated free fallback (Gemini via LiteLLM) auto-catches session/rate limits and retries — limits don't stop work.

discord — a full second front-end

Chat with agents

Per-project channels; the same conversation as the web Chat, mirrored both ways.

Approve in place

Risky actions arrive as Approve / Deny buttons — decide from your phone.

Live board

An #activity feed that edits itself: column counts + who's working on what.

Files both ways

Drop an image → an agent reads it. Agents send files, charts and assets back.

Role-gated

Access is the AgentOS role you hand out — give a friend the role, they're in.

Signals & alerts

Agents post trading signals, digests and status straight to your channels.

the owner surface — what reaches you, and how you steer

Agents have email

Real addresses on the platform's domain: agents send (Resend) and receive (catch-all inbound → their inbox tools) — they read verification codes and complete email-only signups without you.

Owner Actions

When something truly needs a human — an account, capital, a signature — it lands as an action item with its own page: full context, reply to discuss, resolve when done. The agent reads your answer back.

Daily briefing

One morning message in its own channel: yesterday's work, research and spend, plus the few things that need you — each deep-linked to its page. Ask follow-ups in chat.

Decision log & analytics

Every approval decision is recorded with its source and reason; spend by model, project and schedule lives on the analytics page. Audit anything after the fact.

voice (Whisper · Piper)installable PWA + pushwebhooksglobal search ⌘Ksite hostingrecipes export/import
self-improvement & knowledge

Weekly check

Every Monday it researches new models, pricing, skills & trends, self-analyzes the codebase, ships safe improvements, and posts a Discord digest.

Worktree isolation

Self-improvement runs work in an isolated git worktree — they can never touch your live tree or collide with your work.

Agents teach agents

Agents record lessons; the weekly curator promotes durable ones into shared skills — knowledge compounds across every future run.

Debate-gated auto-deploy

Everything auto-ships — even safety-core — after the full suite, a multi-agent debate review (correctness, security, oversight veto, tests + an adversary), a smoke check, and auto-rollback. Core changes loudly notify the owner post-deploy.

safety model

The approval gate

Safe actions run free; sudo, service control, destructive, out-of-workspace writes pause for a one-click decision — in the web or Discord.

Flat permissions

Admin gates one thing: user accounts. Every other capability is open to any authenticated user.

Break-glass recovery

Agents govern themselves — the one unremovable safeguard is an out-of-band owner recovery (boot invariant + SSH script) the platform's own code can never disable. Locked out is impossible.

Three owner touchpoints

The owner is pinged only for results, human-only actions, and real emergencies — a daily briefing, an Owner Actions page (/action/:id) with reply/resolve, agent email (send + inbound on kasmo.dev), and session-limit auto-retry underneath.

argon2id authhttpOnly sessionsHTTPS · Caddysandboxed file accesscommand allowlistappend-only event log
under the hood

Stack

Node + TypeScript · Fastify · SQLite · a React SPA · the Claude Agent SDK. One service serves REST + WebSocket and drives the agents.

Processes

Three under pm2: the app, the Discord bot, and the LiteLLM proxy (fallback engine) — behind Caddy on HTTPS.

Ops

Private GitHub mirror auto-pushed on every deploy — manual and autonomous. Runtime state lives outside the repo; secrets never leave the box.