Agents that remember
Every agent keeps its context between runs and is versioned in Git — it picks up where it left off, and you can roll back.
Trinity is the production runtime for your AI agents — governed, auditable, and inside your perimeter.
1,500+ clones in the last 2 weeks

Agent fleet
Real screenshots from the Trinity control plane. No marketing mockups. The agents pictured are ours.
13 agents, 500 messages in 24h, every one of them health-checked.
Every MCP call, every scheduled run, every failure on one axis. Scrub. Replay.
Cron, timezone, next run, last run, full execution log — visible to operators, not just the engineer who wrote it.
Async by default. Agents run on their own; only escalations land in your queue. Approve, redirect, or fix — in batches, on your schedule.
Get started
One command on your own server — or spin up a managed Cloud instance.
Plug your Claude Code agents in over MCP. They get shared state, identity, and memory.
They run on schedule, coordinate, and every action is audited — inside your perimeter.
In production
“We run our entire production agent fleet on Trinity — multi-user, scheduled, and audited, all inside our own perimeter. It just runs.”
We run dozens of our own agents on Trinity — Cornelius, Corbin, and Ruby are just the three we ship as templates. It is not a side product. We deploy what we sell, and we page ourselves first.
The control plane
Same workflow, same agent files, same git history. Push to a Trinity instance and the infrastructure work disappears underneath.
Every agent keeps its context between runs and is versioned in Git — it picks up where it left off, and you can roll back.
Self-hosted in your own cloud, behind your VPN. Agents run sandboxed; credentials stay encrypted at rest.
Message your agents from Slack, Telegram, or WhatsApp — or wire them to webhooks and your own tools.
The runtime
| Where you are today | Where Trinity fits |
|---|---|
| Use Claude Code if you want a coding agent on your laptop | …those agents need to run in production — off laptops, multi-user, scheduled, audited. (Trinity works with Claude Code via MCP.) |
| Use OpenClaw / Hermes if you want an open coding-agent harness on your machine | …those agents need a production home — hosted, multi-user, scheduled, audited, in your perimeter |
| Use Multica / Paperclip if you want a managed app or board to run agents for your team | …agents are company infrastructure you self-host, govern, and audit in your own perimeter — not a SaaS you send your data to |
The cognitive core
Cornelius is the self-improving cognitive core that lives on Trinity. Pictured is not a mockup — it is Cornelius rendering its own mind: a live 3D graph of notes, edges, and activity that reshapes itself as the system learns your business.
Editions
Open core - Enterprise includes everything in open source.
Enterprise
A dedicated, single-tenant Trinity instance — inside your own perimeter.
Related content
A newsletter for people running AI agents in production on Trinity: hands-on guides, workshop invites, and the releases that actually matter.