AI agent platform with an open-source core
AI agents that show their work. A person who answers for it.
AI makes it easy to produce more: code, content, research, whole projects. The hard part is trusting what it produced. Joyus is built on one rule: a person directs the work, and every step is recorded, so you can check what was done and why instead of taking it on faith.
Two questions sort every piece of AI work
Who made the call, and can you check the result? The work that holds up sits in one corner: a person's call, with a trail anyone can inspect. Joyus is built for that corner.
Autonomous & asserted
The model decides, and you take its word. Fast, and unaccountable.
Autonomous & inspectable
The model decides, but leaves a trail. Better — though no one chose.
Human & asserted
A person decides, but nothing is written down. Trust without proof.
Human judgment, on a receipt
A person makes the call, and every step is recorded so anyone can check it. That record is the receipt. Joyus lives here.
The core is open source. Clone it today.
Saying the work is checkable wasn't enough, so the platform core and Joyus Desktop are public under Apache 2.0. The organization-specific skills, customers' private content, and the commercial services stay closed. We call that open core: the foundation is public and forkable, the commercial layer on top is not, and the repo says exactly where the line falls.
Joyus AI platform
The platform core: MCP (Model Context Protocol) services, session and context state, workflow enforcement, content infrastructure, pipeline orchestration, and the specs that govern those systems.
Joyus Desktop
The native companion app: policy enforcement, session handoff, skill and MCP distribution, managed git sessions, and local-versus-remote environment management.
Products built on the open core. Proprietary, and honest about it.
These run on the open platform but aren't open themselves. They exist and they work — we'll show you. None of them are "go inspect the repo"; they're "come find us."
Joyus Enrichment
Turns Drupal-managed content into reviewable, citable enrichment packages — proposed components, field and taxonomy updates, media, each carrying its provenance and an approval step before anything touches your site. The data model, the module, and a working demo exist; billing and provider wiring are still being finished. Not for sale yet.
Joyus Profile Engine
Voice and authorship profiles: analyzes a corpus to build an author profile, verifies attribution, and monitors fidelity drift. Measured at 97.9% attribution accuracy on a 9-author, 108-document legal corpus, and F1 = 0.935 across 53,000+ Drupal-core comments. We will publish the methodology, so the numbers can be checked, not taken on faith.
Joyus Context
A shared memory substrate — one provider-agnostic way for Joyus tools to store, retrieve, explain, and supersede durable knowledge across sessions and tools.
Joyus Recon
Site architecture discovery and competitive intelligence: architecture reference documents, security reviews, and pre-drafted work packages — from a single public URL through full codebase access.
Joyus Docket
Turns CSV and Jira backlogs into local, agent-ready work queues — triage, planning, critic review, and one-issue commits, with imported data kept as immutable truth and workflow state layered on top.
What we won't pretend
- Open core, not open everything. The platform core is Apache-2.0; the skills, customers' private content, enrichment, and ops are proprietary.
- Proven method, alpha platform. The approach is tested; the platform itself is early.
- Enrichment is early access. It exists and runs, but it isn't for sale yet — so we won't tell you to buy it.
- Numbers carry their scope. Every figure travels with the corpus it was measured on and a methodology we will publish so you can check it.
- A human signs the work. A person directs, reviews, and ships. AI amplifies; it doesn't sign.