The Operating System for the AI-Native Company
Run AI agents 24/7 on your own infrastructure — scheduling, multi-agent coordination, credentials, and monitoring. Open source.









Dashboard
Grid view — the fleet as an org chart: department zones, reporting lines, live tiles
Deploy Trinity in one click
Trinity 1.0.5 is published as a 1-Click Droplet image on the DigitalOcean Marketplace, as an AMI on the AWS Marketplace and as an app on the Vultr Marketplace. Pick a size, launch, and open the address you get — your fleet is running on infrastructure you own. Trinity itself is free and open source under Apache 2.0; you pay the cloud for the machine.
DigitalOcean 1-Click Droplet
The fastest route to a running instance, and the only one with no terminal at any point. 4 GB / 2 vCPU minimum, 8 GB / 4 vCPU for a working fleet.
- First boot in about ninety seconds — the images are baked into the snapshot, so nothing is pulled
- HTTPS at the Droplet's own IP, with a real Let's Encrypt certificate and no browser warning
- You claim the admin account in the browser — no terminal, no console, no password to copy
- Docker, Caddy and a host firewall configured before Trinity starts
Built on Ubuntu 24.04 with Trinity 1.0.5, Caddy and Redis. View the Marketplace listing. Prefer to set the admin password before the Droplet exists? Use the installer script.
AWS Marketplace AMI
Single-tenant, in your own VPC. Subscribe, launch an instance with a public IPv4 and ports 80 and 443 open. Large sizes (2 vCPU / 8 GB, e.g. t3a.large) run the platform; xlarge gives several agents headroom.
- Launch the AMI from your own AWS account — first boot takes a few minutes, nothing to install
- HTTPS at the instance's public IP, with a real Let's Encrypt certificate and no domain required
- You claim the admin account in the browser with the EC2 instance ID, so whoever finds the IP cannot take it
- Docker, Caddy, a host firewall and bundled PostgreSQL ready before Trinity starts; the certificate follows a changed IP after a stop/start
Built on Ubuntu 24.04 with Trinity 1.0.5, Caddy, Redis and PostgreSQL. 64-bit x86. Prefer your own EC2 instance? Use the install script.
Vultr Marketplace app
Deploy from the Vultr Marketplace onto a plan with at least 4 vCPUs and 8 GB RAM — the deploy page does not offer smaller ones — and a public IPv4 address.
- Vultr installs Trinity on a stock Ubuntu 24.04 at first boot — about five minutes, nothing to run yourself
- HTTPS at the server's own IPv4 address, with a certificate issued for that address and no domain required
- You claim the admin account in the browser — the first person to open the address creates it, so open it right away
- Always the latest stable Trinity release, with Docker, Caddy and a host firewall configured before Trinity starts
Runs on Ubuntu 24.04. The listing opens in the Vultr console. Prefer an instance you created yourself? Use the install script.
Choose Your Path
Trinity is the platform that runs agents 24/7. Abilities is the toolkit for building them in Claude Code.
Building Agents
Create agents with Claude Code and the abilities plugins
/plugin install create-agent@abilityai /create-agent:custom /trinity:onboard
Deploying Trinity
One click on DigitalOcean, AWS or Vultr, any Linux server, or your own computer
# DigitalOcean: 1-Click Droplet # AWS: Marketplace AMI # Vultr: Marketplace app # Your own machine: git clone github.com/abilityai/trinity ./scripts/deploy/start.sh --unattended
Using Trinity
Navigate the dashboard, manage agents, set up schedules
# Open http://localhost # Create agents, configure credentials # Set up cron schedules
The Complete Course
Everything in Trinity 1.0, end to end, in one video — jump to any of the 41 chapters from the YouTube timeline.
Trinity 1.0 — the complete course
76 min · 41 chapters
Prefer a guided path? Trinity Learning
Learning tracks by role — everyone, operators and IT, engineers, evaluators and partners — built from short narrated films with quizzes, chaptered courses, workshop recordings and these docs. Invitation-only for now.
Platform Features
What Trinity provides for agent deployment and operations.
Isolated Containers
Each agent runs in its own Docker container with dedicated resources, three agent runtimes (Claude Code, OpenAI Codex, Gemini CLI), and model selection per task.
Fleet Observability
Graph view topology, Gantt-style execution timeline, host telemetry, fleet health monitoring, and OpenTelemetry metrics export.
MCP Integration
129 tools exposed via the Model Context Protocol. Connect Claude Code, Cursor, or any MCP client for external orchestration.
Agent-to-Agent Delegation
Hierarchical delegation with fine-grained permissions, parallel task execution, and shared folders via Docker volumes.
Credential Security
Direct file injection with encrypted git storage, centralized subscription management, and OAuth2 flows for external services.
Operations Dashboard
Real-time monitoring, cron scheduling, live execution streaming, public agent links, mobile PWA admin, and cost tracking per agent.
Videos & Workshops
Platform demos, agent deployments, and live workshop recordings.

Inside Trinity: Agentic Operating System for AI-Native Companies
Trinity explained in plain terms, interview style — how a 10-person company runs on a fleet of agents, the workspace and approval inbox, why a company should own its intelligence, model-agnostic agents, security, and how to install it yourself

Giving My AI Agency Payment Rails and a Finance Agent (Agent-Native Agency Workshop, Session 6)
Session 6, the finale, makes the agency production-ready — a week of unattended schedules in review, a credential convention with one vault per agency, Stripe as payment rails on a restricted key, and a finance agent for invoicing and receivables built from the library

I Let an Agent Make Every Decision on My New Website (Agent-Native Agency Workshop, Session 5)
Session 5 lets real people into the agency and gives it a front door — seats, roles, people and agents; approvals routed to seats in the canon; one agent running a whole website through pull requests; and a website form that becomes a deal in the sales rep's inbox

The Moment My AI Agents Talked Me Out of a Discount (Agent-Native Agency Workshop, Session 4)
Session 4 turns separate agents into one system that runs itself — an orchestrator that routes work, tracks receipts and handles escalations, and the Canon every agent reads and writes; live on a ~40-agent instance, then built by hand in the Keymaker agency

The Open-Source Second Brain for Your Company (Agent-Native Agency Workshop, Session 3)
Session 3 gives the agency a brain — Cornelius, a company second brain installed from a template inside your own Trinity: first-run indexing, how knowledge gets in and how it is retrieved, and agents collaborating with the brain live

Staffing an Entire Agency With AI Agents (Agent-Native Agency Workshop, Session 2)
Session 1 stood up the instance; this one staffs it — sales, marketing, and delivery agents installed from the Trinity library, a live sales agent run, an existing Claude Code agent converted and deployed from GitHub, and permissions wired so four agents talk to each other

40 Agents, One Instance (Agent-Native Agency Workshop, Session 1)
A tour of the production instance running ~40 agents — the org chart, who may talk to whom, credentials, approval chains, per-person memory — then a Trinity install on DigitalOcean from zero, live

Make Your Agency AI Native (Webinar)
The opening session of the Agent-Native Agency series — a live look inside a production fleet of ~40 agents organized as an org chart, and the framework for turning agents that already run in silos into one system that compounds

Trinity v0.9.0 — Release Tour
Multi-agent org structures on the Grid, the Workspace in every build, the shared skills library, and bidirectional A2A — a walkthrough of the v0.9.0 release

How I Run My Company with a Fleet of 30 AI Agents
A hierarchical agent fleet that works like an org chart — orchestrator, managers, workers — with a shared canon, a permission matrix, GitHub-issue project management, and human approval gates

Make Claude Code Agents Run Autonomously on Trinity (Live Workshop)
Analyze, standardize, and migrate a pile of home-grown agents into a managed fleet with the agent-dev plugin, then deploy one from GitHub onto a live Trinity instance

How to Build a Second Brain for Your AI Agents (Live Workshop)
The 5 rules for managing agent knowledge — Decisions vs. Facts vs. Judgment, knowledge as code in Git, a central mind-center agent, and human-endorsed canon

Cornelius AI Agent — Architecture Explained
How a second-brain agent turns raw insights into a typed knowledge graph, reasons autonomously to convergence, and never publishes without a human gate

Trinity v0.8.0 — Release Tour
Voice replies across channels, the Brain Orb, the Grid dashboard, and the enterprise identity layer — a walkthrough of the v0.8.0 release

Measurably Self-Improving Multi-Agent System Without Retraining
Brier HQ — a forecasting fleet on Trinity that improves as it runs: five feedback loops, 96,161 predictions, zero retraining

One Agent Per Client: A Scalable AI Service Platform
The end-to-end pattern for agencies — deploy sovereign, isolated agents per client, each with their own data, credentials, and channels

Cornelius & the Brain Orb
A second-brain agent renders and talks to its own 3D knowledge graph — live

Build Autonomous Loops for Your AI Agents
Loops engineering with Trinity — agents that repeat bounded tasks with templates and stop signals

I Gave My AI Agent an Unconscious Mind
A live build of Cornelius into a fully autonomous thinking agent on Trinity

Production-Grade Hosting for the Agents You've Already Built
Full Trinity demo — deploy the agents you've already built to production-grade hosting

The Multi-Agent Platform I Run My Company On
A live tour of Trinity — the multi-agent platform that runs Ability.ai

Trinity Ops — the Agent That Deploys Other Agents
A live walkthrough of the public Trinity Ops agent

Why Every AI Agent Needs a GitHub Repo
Using GitHub as the backbone for autonomous agents

3 AI Agents Run My Entire Software Development Pipeline
The Dev, Ops, and Docs agents managing the Trinity platform in production

Plugins + Trinity: Build and Deploy Agents in Cursor
The complete loop from empty folder to deployed autonomous agent

Building Agents — Playbooks, Plugins, Deployment
A deep dive into Playbooks and the Playbook Builder

Your First Trinity Agent in 30 Minutes
From zero to a working agent — setup, configuration, and deployment
Explore the Documentation
Reference documentation organized by topic.
Getting Started
Three paths: build agents with Claude Code, deploy Trinity, or learn the UI.
Architecture
System design, multi-agent coordination, and agent lifecycle.
API Reference
REST API endpoints for agents, chat, schedules, and webhooks.
Guides
Step-by-step instructions for agents, credentials, scheduling, and integrations.
Cloud Code Plugins
Claude Code plugins for agent creation, development, and Trinity integration.
Cornelius
The open-source second-brain agent built for Trinity — a company or personal brain that compounds, with receipts.