Stop renting agents.
Start owning them.
NOMOS lets you buy autonomous AI agents once and run them forever — locally, on your machine, powered by ChatGPT, Claude, Gemini, or any local model. No per-action fees. No platform lock-in.Your data never leaves your computer.
NOMOS Exchange
The first true open app store for the agentic economy. Don't build plugins for a corporate walled garden. Browse, publish, and monetize independent agents that run entirely locally.
Desktop Hypervisor
Connect ChatGPT, Claude, or Gemini and let your agents run 24/7 in the background — automating tasks while you focus on what matters.
NOMOS SDK
Build agents in Python that work with any LLM — ChatGPT, Claude, Gemini, or local Ollama models. VS Code extension included: IntelliSense, live trace viewer, and NomosUI snippets. Publish once, reach every NOMOS user.
Powered by the models you already use
No API key? Start in 90 seconds anyway.
NOMOS auto-downloads a free local AI model that runs entirely on your hardware — great for trying agents before connecting your preferred subscription. Switch to ChatGPT, Claude, or Gemini any time in Settings.
Decoupled Intelligence
NOMOS separates intelligence from infrastructure using three core protocols.
MAA
Swap between cloud and local models (Ollama, OpenAI, Anthropic) without rewriting logic. Seamlessly transition as models evolve.
MCP
44+ integrations including Gmail, Slack, GitHub, Notion, Stripe, Jira, and more. One-click OAuth2 connections. Credentials never leave your NOMOS environment.
A2A
Agents recruit specialists by name or capability tag — route to whoever handles the job, not just who you know.
Your Personal AI Workforce
Create a named team in the desktop UI. Add your specialist agents. From that point, they discover and delegate to each other automatically — scoped only to teammates you chose.
Complex problems,
simple instructions.
Agents within the same team automatically discover each other. When a task exceeds one agent's scope, it delegates to a teammate — passing context, waiting for results, and assembling the final answer.
Team-Scoped Discovery
Agents find teammates by name or capability tag within the team you configured — discovery never leaks outside the team boundary.
Natural Language Delegation
Instructions are plain English. The receiving agent's LLM interprets and executes.
Loop Prevention
Built-in recursion guard — specialists cannot re-delegate, preventing infinite chains.
Local-Model Aware
10-minute delegation timeout accommodates slow local inference end-to-end.
Sovereignty is Choice.
Everything runs on your machine. Your conversation history, your API keys, your agent logic — none of it touches a server you don't own.
That's not a feature.
That's the architecture.
Autonomous — not reckless.
When an agent is about to do something irreversible, it stops and waits for your approval. Every time. No exceptions.
Action Requires Approval
agent · file-organiser
You stay in control.
Always.
Agents can think and plan autonomously, but irreversible actions are always gated. You see the exact tool call arguments before anything is executed.
File & System Operations
Delete, overwrite, move — any file system mutation triggers the gate before execution.
Outbound Communications
Email sends, Slack messages, API writes — agents can draft, but you decide when to send.
Full Argument Visibility
See the exact parameters the agent intends to pass — recipients, paths, content — before approving.
SDK-Level Enforcement
The guard is baked into the NOMOS SDK, not a UI wrapper. Agents built on NOMOS cannot bypass it.
Your agent wakes up so you don't have to.
A payment lands, a file appears, a deadline hits — your agent is already on it. No polling, no checking, no manual triggers.
Instant Connections
When a payment arrives, a form is submitted, or a service sends an alert — your agent knows immediately and takes action, before you even see the notification.
Drop a File, Get It Done
Drop an invoice, a report, or a photo into a watched folder and your agent picks it up automatically — processes, files, and responds without you clicking a thing.
Runs on Your Schedule
Set your agent to run every morning, every weekday at 9:30, or on the last day of the month. Configure it once in the app — it just happens, reliably.
Agents Work Together
Agents coordinate behind the scenes — one finishes a task, another picks up where it left off. Complex multi-step workflows, zero manual handoff.
Start it. Forget it. Come back to results.
Your agent runs long jobs in the background — while you work on other things. No timeouts. No babysitting. Just results waiting for you.
Survives Restarts
Jobs persist to disk. A crash or restart doesn't mean starting over — the agent picks up where it left off.
Real-Time Progress
Watch your job inch toward 100%. Know exactly what it's doing, how far along it is, and what comes next.
Automatic Retry
Transient errors don't kill the job. Configure retry limits and exponential backoff — the agent handles the rest.
Cancel Anytime
Change your mind mid-run. The agent stops gracefully at the next checkpoint — no stuck processes, no cleanup required.
Agents that ask — not just tell.
When your agent needs a decision, a date, or a preference — it asks properly. No more typing instructions into a chat box.
Ask for What It Needs
Drop-downs, checkboxes, file uploads — agents collect exactly what they need in a form you actually recognize.
See Your Data, Your Way
Kanban boards, calendars, timelines, tree views — information in the format that actually makes sense for the task.
Work With It, Not Around It
Click to edit a cell. Drag a card to a new column. Delete a row. Your agent's data is alive.
Guided Step by Step
Tabbed views and step-by-step wizards keep complex tasks organized — no scrolling through walls of text.
Results delivered — wherever you are.
Your agent doesn't wait for you to open the app. It sends the report, files the document, and pings you — through every channel you already use.
Desktop Alerts
Get a native notification the moment your agent finishes — even when the app is in the background or your screen is locked.
Email, Slack & More
Receive results directly in Gmail, Slack, or Discord. Your agent sends the message — you just read it, wherever you already are.
Ready-to-Open Files
Reports, spreadsheets, PDFs — saved automatically to your Outputs folder with a single click to open. No copy-pasting, no manual exports.
Live Dashboard
A live summary view that updates every time your agent runs — charts, tables, key numbers — always current without you doing anything.
Chat from Anywhere
Connect via Telegram to securely access your local agents remotely. No exposed ports, no webhooks—just long polling with security whitelist.
24/7 Background Autonomy
Agents run silently in the system tray with a 60-second heartbeat. Lock-aware mutexes prevent race conditions.
Use cases: Monitor inbox for urgent emails, check RSS feeds, alert you to anomalies—all while you sleep.
Agents that remember everything.
Built-in storage primitives so agents carry context across every restart, run, and conversation.
KV State Store
set/get arbitrary values across every run. TTL support, prefix scoping, and atomic update().
Time-Series Metrics
Track counters and rates. Query by window ("24h", "7d") with sum/avg/min/max aggregation.
Shared Team State
Gateway-backed KV shared across all agents on a team. Zero infrastructure — built into the runtime.
Context Memory
Long conversations auto-summarize instead of truncating. Old history compressed, never lost.
Agents that orchestrate agents.
Chain, fan-out, and monitor entire teams of agents with a single API — no custom coordination code required.
DAG Pipelines
Chain agents with dependency resolution. Independent stages run in parallel; downstream stages wait for their deps.
Fan-Out / Fan-In
Broadcast the same task to N agents simultaneously and collect results with .successful() and .first_success().
Agent Supervisor
Monitor team health in real-time. Register async callbacks on agent failure — no polling, no glue code.
Structured Results
PipelineResult carries per-stage results, errors, and a final_result. Context flows automatically between stages.
Know what you need before you click Install.
Every agent shows its requirements upfront — which keys are set, which integrations are connected, and what's missing — so there are no surprises after install.
See Before You Install
A requirements checklist appears before you confirm any install. LLM keys, env vars, OAuth integrations — all checked and displayed instantly.
Readiness Score
Every agent in your Library shows a 0–100% readiness score. At a glance, see which agents are fully configured and which need attention.
Never Blocked at Launch
Missing a key? Still install it. The checklist is informational, not a gate. Configure missing items in the Config tab at any time.
Guided Configuration
A step-by-step wizard walks you through every required setting before the first run — no cryptic variables, no startup failures.
Build Once. Earn Forever.
Build an independent business, not a cloud-locked feature. NOMOS handles licensing, distribution, and payments on a protocol you actually own.
Set Your Price
One-time or subscription. You decide. NOMOS Credits (NC) power all verified purchases.
Real-Time Dashboard
Track active items, download metrics, and revenue analytics in your creator portal.
Fair Revenue Share
Transparent splits. Instant payouts. No gatekeeping—publish what you build.
Atomic transaction system prevents license duplication. Creator ownership bypass—your agents are always yours.
The Exchange
isn't open yet.
We're onboarding the first cohort of NOMOS builders — developers who want to publish agents on the Exchange before it opens to the public.
First cohort · Limited spots