Skip to main content

The operating system for AI-native companies

Run AI agents inside your own infrastructure, with permissions, shared memory, and a full audit trail. Cornelius, its cognitive architecture, turns their work into knowledge your organization keeps.

Trinity is the open-source (Apache 2.0) operating system for AI agents, with Cornelius, its cognitive architecture, built in. It is a self-hosted AI agent platform to run agents in production - a runtime built by Ability AI that runs Claude Code, Codex, and Gemini CLI agents inside your own perimeter with git-versioned state, multi-agent orchestration, governance, and an append-only audit trail. Install it from GitHub, in one click from the DigitalOcean Marketplace, or from the AWS Marketplace.

Inside your perimeter.

Trinity deploys on hardware you own — a workstation, a rack, a VM behind your VPN. Models, keys, memory, audit trail: all of it stays inside. The platform never faces the internet; the only public surface is the channels you choose to open. Your infrastructure, your data, your agents — owned, not rented. Your competitors rent the same AI; this is the version you own.

Bring the agents you already build.

Claude Code locally is a developer tool. Trinity is a runtime. An agent is a git repo of prompts and skills — Trinity runs Claude Code, Codex, and Gemini CLI agents side by side. Same repo, same skills, no rewrite.

Its own sandbox. Its own repo.

Every agent runs in an isolated container and commits its memory back to git every 15 minutes. State accumulates - every decision, note, and fix stays in the agent's history, closer to an employee than a script. The more a system improves itself, the more it matters who owns it.

They work as a team.

Agents message each other, delegate, and hand off — research to draft to outreach — only across permissions you've granted. Multi-step workflows run end to end with no human relaying context between steps.

A mind at the center.

Cornelius is the cognitive architecture at the center of the fleet - a living knowledge graph that accumulates your company's entities, decisions, and hard-won insight, and serves them back to every agent. The longer it runs, the more of your company it holds.

Full control from your favorite AI.

The whole platform is exposed as an MCP server — 90+ tools. Connect Claude, ChatGPT, or Cursor and run the fleet in plain language: create agents, trigger runs, read logs, wire schedules, rotate credentials. Everything the dashboard can do, your AI can do.

Who can do what — enforced.

Permissions for users and agents alike: who can see, who can call, who can approve. Guardrails cap tools, time, and budget per run. Every action lands in an append-only, hash-chained audit trail no one can quietly edit.

It runs on schedule — and pauses for you.

Cron schedules, loops, event triggers, pipelines — and fan-out, one task dispatched across many agents at once. A 3 a.m. run fires unattended; the one irreversible step waits at a human gate. You approve from your phone over coffee, and the run resumes exactly where it paused.

Meet them where your team lives.

Expose agents to your team — or your clients — over Slack, WhatsApp, Teams, or Telegram. One identity gate on the perimeter; the platform itself stays behind the VPN.

Build once. Deliver to clients.

Every agent you build is a portable git repo. Deliver it as a service — clients reach it through a gated channel or a portal, with exactly the permissions you grant. Or install it inside their perimeter, on their own Trinity, where the client owns the install. What you build for one client compounds into the next.

cornelius avatar
corneliusSYSTEM
acme/cornelius
Running
▶ consolidating graph12,408 nodes
Activity · 14d—
Context · 7d—
Success97%$41.20 · 2m ago
chief-of-staff avatar
chief-of-staff
acme/chief-of-staff
Running
cron 07:00▶ morning brief
Activity · 14d—
Context · 7d—
Success92%$12.40 · 14m ago
marketing avatar
marketing
acme/marketing
Running
▶ drafting campaign2 in queue
Activity · 14d—
Context · 7d—
Success88%$9.85 · 41m ago
sales avatar
sales
acme/sales
Running
⏸ gate · email.send31 drafted
Activity · 14d—
Context · 7d—
Success84%$18.02 · 3m ago
customer-support avatar
customer-support
acme/support
Running
▶ ticket #4312queue 3
Activity · 14d—
Context · 7d—
Success96%$22.60 · just now
market-research avatar
market-research
acme/research
Running
cron Mon 08:00watching 14 rivals
Activity · 14d—
Context · 7d—
Success90%$7.44 · 1h ago
back-office avatar
back-office
acme/back-office
Running
▶ CRM hygieneinbox zero 2d
Activity · 14d—
Context · 7d—
Success81%$11.90 · 22m ago

Open source is the whole platform. Enterprise adds governance.

Open source — Apache 2.0, free

  • Agent platformfleet, chat, voice, web terminal, shared workspace
  • Multi-runtimeClaude Code, Gemini CLI, OpenAI Codex
  • Orchestrationscheduling, agent loops, agent-to-agent calls, recovery
  • System agentbuilt-in ops agent: health, restarts, schedules, costs
  • Channels & APIsSlack, Telegram, WhatsApp, voice calls, MCP, REST, CLI
  • Git-native agent stateGitHub sync, commit history, reset to upstream
  • Marketplace & librariesplugins per agent, skills library, templates
  • Security coreencrypted credentials, hardened containers, audit log
  • Guardrails & cost controlssafety rules, read-only mode, quotas, loop budgets
  • Tamper-evident audit logopt-in SHA-256 hash chain with verification
  • Roles & sharingfour roles, per-agent sharing
  • Self-hosted deploymentyour hardware, no telemetry unless you opt in

Enterprise adds

  • User & org managementinvite, deactivate, reactivate, activity per user
  • Single sign-on (OIDC)Okta, Entra ID, Google Workspace
  • Two-factor authenticationauthenticator apps, recovery codes, policy per role
  • SAML 2.0 SSOSAML single sign-on
  • SCIM provisioningautomated user lifecycle from your identity provider
  • Advanced RBACteams, granular roles
  • Fleet permissions matrixone view of which agent may call which
  • Compliance suiteaudit dashboard, SIEM webhook export, auditor export
  • Managed data retentionconfigurable retention windows beyond the defaults
  • Credential vaultnamed encrypted secrets, deny-by-default grants per agent
  • A2A managementexposure control, inbound allowlist, outbound endpoints
  • Cross-model validationa second model reviews an agent's result
  • Advanced DevOps agentoperates your Trinity instances
  • Ephemeral agentsshort-lived agents with a lifetime and budget limit
  • License managementoffline-verified, air-gap friendly

Open core — Enterprise includes everything in open source.

Deliver agents to your clients? Become a Trinity partner.

One-click · DigitalOcean Marketplace

Deploy it in five minutes.

Pick a droplet and the platform brings itself up — Trinity, Caddy, and Redis on Ubuntu 24.04, inside your own DigitalOcean account. Nothing to configure, nothing of yours passing through us.

  1. Hit “Create Trinity Droplet”. Pick a region and a size — the image is already chosen for you.
  2. Give it 90 seconds. First boot installs the stack, closes everything but 22/80/443, and issues a Let’s Encrypt certificate — no domain required.
  3. Open the HTTPS address. Claim the admin account in the browser and your fleet is live.
Deploy on DigitalOcean →

Ubuntu 24.04 · Trinity 0.9.5 · 4 vCPU / 8 GB recommended · your account, your bill, your data

Not on DigitalOcean? Launch it in your own AWS account from the AWS Marketplace, or clone it onto a Mac mini in the closet, bare metal in your rack, any other cloud — Docker Compose and one script.

git clone https://github.com/abilityai/trinity

The full install guide · single-server, hardening, upgrades

Dive deeper

Watch Trinity in motion.

Trinity Platform — architecture explainedDeep dive · Eugene Vyborov
Production-grade hosting for the agents you've already builtDemo · Eugene Vyborov
Trinity v0.9: multi-agent org structures, Workspace, shared skills & A2ARelease · Eugene Vyborov

Just shipped: the latest release notes - everything in the release, in writing. Running it already? star it on GitHub - it is how other people find Trinity.

Prefer reading? Best AI agent platforms to run agents in production — the flagship guide.

View on GitHub★ 639