The local engine on Linux x86_64
The openeng binary — a Rust engine that runs only on your machine, binds no inbound port, and owns the agent loop, models, memory, and every cloud call. Live and installable today on Linux x86_64.
General availability · Q4 2026The launch is close.
The launch is close.
The engine is already real.
OpenEng reaches general availability in Q4 2026. But the core isn't a promise — the local engine, the browser Web Console, and 3,300+ composable building blocks are live and tested today. Join the waitlist for launch-day access and updates.
No spam, no telemetry — one note when launch-day access opens. Submitting opens your email client to send us a quick note; or email us directly, or start now below.
Open the Web Console
Start your engine
openeng serveNot a roadmap — what's already shipped
1,092Agents
885Skills
754Templates
28On-device models
550Cells
44Loop strategies
2,433Engine tests passing
Honest status
Shipped today, and what's next.
We keep a hard line between the two. Everything on the left is live and verifiable right now; everything on the right is on the roadmap to launch. No vaporware in the 'shipped' column.
Shipped today
Live and verifiable right now
The Web Console, end-to-end encrypted
A thin React client that drives your own engine over an end-to-end-encrypted channel — even the broker that relays it can't read a byte. Your open-weight models run on your machine.
8 catalogs · 3,300+ building blocks
Agents, skills, templates, cells, models, plugins, and loops — browse them, snap them together, and publish a config your engine just runs.
- 1,092 agents · 885 skills · 754 templates
- 550 cells · 28 models · 44 loop strategies
Complexity routing — no modes
An embedded model scores each request 0–100 before spending an LLM call, then self-routes plan-vs-execute. You never pick 'chat vs goal vs coding.'
A 2,433-test green suite
The engine ships behind a fully green test suite — the model runtimes, the privacy guarantees, the encryption, and the agent loop are all covered and passing.
One account, everywhere you sign in
One OpenEng identity across the Web Console, engine, and Desktop — one-click browser sign-in with a device-code fallback, plus API keys for servers.
On the roadmap
Planned for and after launch
macOS & Windows builds
The engine runs on Linux x86_64 today. Native macOS (Apple Silicon) and Windows x64 builds are in progress — the install scripts already detect platform and will go live the moment each binary publishes.
Full per-worker OS isolation
Kernel sandboxing already confines shell commands and the daemon process. Extending hard OS-level isolation to every individual worker is the next hardening milestone.
Embedded media generation
Today's on-device models cover chat, coding, planning, review, and grounding. On-device image and media generation, sandboxed like every local model, is on the roadmap.
Broader rollout
GA in Q4 2026 opens the waitlist, the desktop apps, and the remaining catalog domains. We ship the pieces as they're proven — not before.
Engineering products (Terminal, Kubernetes, Git, Data, Chat) are a separate suite with their own launch — they aren't covered here.
You don't have to wait
The shipped half is enough to start.
You can run the engine and drive it from the browser today — no waitlist required. The waitlist is for launch-day desktop access, the macOS/Windows builds, and the GA announcement.
step 1 · local
Install the engine
One command on Linux x86_64. It downloads a default open-weight model from the Hugging Face Hub and starts the engine — models from the Ollama library, ModelScope, or a direct URL plug in the same way, no account needed to run it.
step 2 · browser
Open the Web Console
Sign in once at app.openeng.ai. The console connects the instant your engine comes online — through OpenEng's broker, nothing to paste.
step 3 · launch
Join the waitlist
Get the GA note, the desktop builds, and launch-day access — or just see the pricing first.
The install one-liner
openeng.ai/engine/install.shInstall the engine
curl -fsSL https://openeng.ai/engine/install.sh | bashBe there on day one.
Join the waitlist for launch-day access and the GA announcement — or skip the line and start running your private agent right now.