Open Swarm — a governed agentic operating layer

Agentic AI you can actually run.

Open Swarm (oshal) turns agentic AI from one clever prompt into an operated platform: specialized bots on the harness and model that fit their job, thousands of connectors brokered per user, isolated tenants, a workflow generator, and real applications people can open today.

See how it works The trading record Browse applications

Per-bot modelsEach bot chooses its harness, provider, and model

1,300+ connectorsImportable on demand from the marketplace

Tenant + user isolationPostgres RLS and per-user storage

Workflow generatorDescribe work, get a runnable queue

Screenshot of the oshal applications dashboard with manifest-loaded app surfaces — The applications dashboard: every app is a manifest loaded onto one runtime — open a surface or toggle bots without restarting.

Screenshot of the oshal assistant Jarvis with voice controls and command center — Jarvis: a single front door that routes voice and text into the right app and workflow.

What's new

Four proofs the platform is real, not a slide deck.

The platform now carries a live trading book, a connector marketplace, a workflow generator, and a fleet of working applications. Each links to its own page with screenshots and detail.

Live · paper account

Autonomous trading desk

A swarm runs a paper book end to end: signals, risk, fills, and a daily AI-generated video recap. Every trade is tied to the signal that justified it.

See the record & video -> Marketplace

1,300+ connectors

A connector runtime plus a marketplace that imports public API specs on demand, so a bot gets the right approved data and actions instead of generic internet access.

How connectors work -> Generator

Workflow generator

Describe an outcome and the platform publishes a runnable workflow as a ticket queue, with build, ticket, and video pipelines as first-class shapes.

See the workflows -> App fleet

Working applications

Intelligent Operations, Intelligent Workforce, an autonomous trading desk, World Intelligence, Little Monsters, Presentation Studio, travel, shopping, and more — all on one runtime.

Browse the fleet ->

The platform

oshal is Open Swarm: any harness, API, or LLM.

oshal is a multi-tenant, multi-user, multi-agent runtime where specialized bots run different agent harnesses, API providers, or LLMs, coordinate over a Redis Streams mesh, and register new apps, tools, agents, and workflows while the swarm is running.

The point is practical: turn agentic AI into a governed operating layer with tenant and user boundaries, tickets, roles, connector-backed data access, cost attribution, review gates, and applications people can actually use.

TypeScript Node.js Redis Streams Postgres + RLS ChromaDB RAG Docker Compose Kubernetes OIDC / Keycloak Connector broker A2A remote agents Local LLMs

Open the architecture ->

Built by Roger Murphy.

Agentic Federal is the proof bench; Open Swarm is the flagship. If you are evaluating the platform, or hiring for senior AI platform, full-stack, or automation work, the About page has the engineering story and a resume.

About Roger Email