OpenEng Labs

OpenEng Engineering · openeng.app

Your engineering workbench, in the browser, on your machine.

OpenEng is a suite of five professional developer apps — Terminal, Kubernetes, Data, Git, and Chat — that run in your web browser but execute on a small native engine you run on your own machine.

Five apps in betaFree, every feature
Open openeng.app (opens in a new tab)Talk to us
engine · 127.0.0.1sealed per frame
k8s.openeng.app
ctx: prod-euns: paymentswatch
NameReadyStatus
api-7d9c4f1/1Running
worker-5b8f2a2/2Running
db-migrate-k20/1Pending
terminal.openeng.app
$ cargo test --workspacetest result: ok. 214 passedexit 03.4scopy

Status, as of September 25, 2026

  • Beta

    Five apps in beta, free

    Terminal, Kubernetes, Data, Git, and Chat. Every app and every feature is free.

  • Beta

    Three engines published

    Terminal, Kubernetes, and Data engines run on Linux and macOS, and on Windows through WSL. Git and Chat engine builds are not published yet.

  • Live

    One account for every app

    Sign in once at auth.openeng.app; each app with an installed engine (Terminal, Kubernetes, and Data today) connects to it automatically.

The five apps

Five apps. One account. Your machine does the work.

Each app pairs a browser interface with its own engine on your machine, and every user-facing feature of the five is specified in one of 151 end-to-end automation suites.

Built on and works with

llama.cpp · GGUFOllamaHugging Face HubONNX RuntimeMCPKubernetesHelmPostgreSQLCockroachDBMySQL · MariaDBSQLiteRedisNeo4jMongoDBApache Kafkalibgit2tree-sittergRPCRustReact

How an app works

Private by architecture, not by policy.

The browser draws the interface and a small native engine on your machine does the work, so your code, repositories, clusters, and databases never pass through an OpenEng server.

A thin browser UI talks to a native engine on your machine over a link that is sealed per frame. The engine keeps your shell, kubeconfig, repositories, and connection strings on the machine, and calls OpenEng only at api.openeng.app for sign-in and its TLS certificate, at releases.openeng.app for the update check, and at openeng.app for the update download.Browser (thin UI)draws the interfacesealed per frameloopback onlyNative engine on your machine127.0.0.1shellkubeconfigreposconnection stringsstays on the machinesign-in, cert, updatesapi.openeng.app · sign-in, TLS certreleases.openeng.app · update checkopeneng.app · update download

  • The engine runs on your machine

    Each app is a Rust gRPC engine on 127.0.0.1. It reads your shell, kubeconfig, repositories, and databases where they already are.

    127.0.0.1 · gRPC and gRPC-Web

  • Sealed per frame

    After the session handshake, every call between the page and the engine is sealed frame by frame, and the engine listens on loopback only.

    X25519 · HKDF-SHA256 · ChaCha20-Poly1305

  • Only sign-in and updates reach OpenEng

    The engine calls OpenEng only to sign in and update: api.openeng.app for sign-in and its TLS certificate, releases.openeng.app for the update check at every boot, and openeng.app for the checksum-verified download.

    api.openeng.app · releases.openeng.app · openeng.app

One account, one engine per app

Sign in once. Every app starts its own engine.

One OpenEng account, created with Google at auth.openeng.app, signs you in to all five apps. Each app then pairs with its own engine on a fixed loopback port, so a Kubernetes session never shares a process with your shell or your databases.

One OpenEng account

Google sign-in

auth.openeng.app

  • Terminal

    127.0.0.1:50052

  • Kubernetes

    127.0.0.1:50054

  • Data

    127.0.0.1:50057

  • Git

    127.0.0.1:50053

  • Chat

    127.0.0.1:50056

Each engine binary and the fixed loopback port it listens on.

How you start

From sign-in to a connected engine in four steps.

  1. Step 01

    Sign in once

    Create one OpenEng account with Google at auth.openeng.app. It signs you in to all five apps.

    auth.openeng.app

  2. Step 02

    Install and pair

    Install the engine with its one-line script, then run the pairing command the app shows. It carries a single-use sign-in key.

    engine/<app>/install.sh

  3. Step 03

    The engine starts

    It checks for a verified update, exchanges its key with api.openeng.app, and serves gRPC and gRPC-Web on 127.0.0.1.

    127.0.0.1 · ports 50052–50057

  4. Step 04

    The page connects

    The browser tab finds the engine and connects over a sealed channel. Your shell, clusters, repositories, and databases appear where they already live.

    sealed per frame

Use scenarios

What a team can do with the suite today.

Written from the product documentation, not from customer stories. Each scenario uses only what a team can install today: the Terminal, Kubernetes, and Data engines and the ecosystem connectors.

  • Triage a failing rollout

    Tail and search the pod logs, follow live events, open a shell inside the pod, and port-forward to it, all from one tab. Credentials come from your kubeconfig and never pass through OpenEng.

    k8s.openeng.app

  • Run one command across many sessions

    Fan a command across sessions in parallel, keep folder-scoped history ranked by use, and open incognito groups that never touch disk, all in real PTY shells on your machine.

    terminal.openeng.app

  • Query every data store from one tab

    Connect directly to PostgreSQL, MySQL, Redis, Neo4j, MongoDB, and more, keep connection strings on your machine, and reach RDS, S3, or Kafka through a read-only connector in your own cloud.

    data.openeng.app

  • Live

    Expose your own cloud, read-only by default

    Deploy a source-only connector behind your own API gateway, keep it read-only with a write-key gate, and give the Terminal, Kubernetes, and Data apps its tools. 13 of 252 connectors are validated live.

    connectors.openeng.app

The AI platform

The OpenEng AI platform: a separate launch.

OpenEng is an agentic ecosystem for open-weight model execution: it downloads, orchestrates, and runs open-weight (open-source) AI models entirely on your machine.

Status, as of September 25, 2026

  • Live

    AI engine and Console

    The engine is live on Linux x86_64. The Console publishes an immutable id@version and stores secrets by reference only.

  • Planned

    macOS and Windows

    Engine builds for both will follow the Linux x86_64 build.

  • Planned

    General availability

    Planned for Q4 2026, when the paid plans will launch. The engine is free to run today.

The engine downloads open-weight models and runs them on your machine; no API key is needed for the 28 models in the on-device catalog. Besides its core runtimes, it runs vLLM, SGLang, ExLlamaV2 and V3, MLX, TensorRT-LLM, and Transformers when their toolchain is installed.

The Web Console at app.openeng.app drives your own engine over an end-to-end-encrypted channel. The Console at console.openeng.app is where you author models, agents, templates, execution flows, and cells, then publish them as one immutable id@version.

Core runtimes

  • Embedded GGUF

    llama.cpp, bundled

  • ONNX Runtime

    bundled

  • Ollama

    your local Ollama

  • Network endpoint

    OpenAI-compatible

Web Console
app.openeng.app (opens in a new tab)
Console
console.openeng.app (opens in a new tab) (sign-in required)
console.openeng.app

config

payments-triage

payments-triage@1.3.0 · published

Publish 1.4.0
chat model
qwen2.5-coder-7b
runtime
llama.cpp · GGUF
database url
${PAYMENTS_DB_URL}
cells
aws-rds · read · 1 tool

Secrets are stored by reference. The Console never holds a value.

Model axesThree separate choices; the format decides which runtimes can run a model.
registry
Hugging FaceOllama libraryModelScopedirect URL
format
GGUFSafetensorsONNX
runtime
llama.cppONNX RuntimeOllamavLLMTransformers

  • Secrets by reference only

    The Console stores ${ENV} references, never values. It never runs your agent and never holds a secret.

    ${PAYMENTS_DB_URL}

  • Publish once, strictly higher

    A published config is an immutable id@version, and every later publish must carry a strictly higher SemVer.

    payments-triage@1.4.0

  • Plugins at every call

    14 plugins gate, transform, observe, or enrich a call above a platform floor that no config can remove. Built-ins are free; external plugins will come with the paid plans.

    plugins.openeng.app

  • Agents, roles, and loops

    1,092 agents, 1,524 behavioral roles, and 6 runnable global loops, published as content-as-code. Behavioral roles are a Pro and Enterprise feature and will arrive with the paid plans.

    agents · roles · loops

Content-as-code

3,300+ agents, skills, templates, models, cells, and plugins.

Every catalog is a set of validated definitions, published as content-as-code and served as catalog.json. A published config freezes the ones it uses into its immutable id@version. Roles, library bundles, loops, and connectors are counted separately.

Counts read from each catalog.json on September 25, 2026; the six catalogs add up to 3,323, shown as 3,300+. The 28 on-device models range from 1B to 70B parameters, and the Kubernetes connector alone exposes 151 tools.

Deployment and status

What runs where.

The interface is served from openeng.app and the engines run on your hardware. The only shared services are sign-in, the engine certificate, and the release registry and download host the engines update from. Status as of September 25, 2026.

Where each OpenEng Engineering surface runs, with its platforms and status
Surface
App interfacesBetaServed from terminal, k8s, data, git, and chat.openeng.appAny modern browser
App enginesBetaYour machine, on 127.0.0.1; installed by a per-app scriptTerminal, Kubernetes, Data: Linux, macOS; Windows via WSL
Engine updatesLivereleases.openeng.app, at every boot; Terminal, Kubernetes, and Data also check every 30 minutesDownloaded from openeng.app, checksum-verified before install
Sign-inLiveauth.openeng.app; the engine exchanges a one-time key with api.openeng.appGoogle sign-in
Data cloud connectorsBetaA source-only connector Lambda deployed in your own cloudRDS, S3, DynamoDB, GCS, Kafka; strict read-only mode
Ecosystem connectorsLiveSource-only Python tool servers you deploy in your own cloud13 of 252 validated live; 239 specification only
DocumentationLivedocs.openeng.app, with the AI platform at docs.openeng.app/aiThe five apps, Connectors, and the AI platform
AI engine (separate launch)LiveYour machine; driven from the Web Console at app.openeng.appLinux x86_64 today; macOS and Windows builds planned
ConsoleLiveconsole.openeng.app, sign-in requiredPublishes an immutable id@version; secrets by reference
CatalogsLiveTen catalog.json feeds, from agents to connectorsPublic; no sign-in needed
Paid plansPlannedPro and Enterprise, launching with general availability in Q4 2026Schedulers, the headless runner, and external plugins arrive with them

Directory

Every OpenEng Engineering surface, in one place.

FAQ

Questions about OpenEng Engineering

Which apps are available, and are they free?

Five apps are in public beta today: Terminal, Kubernetes, Data, Git, and Chat, at terminal, k8s, data, git, and chat.openeng.app. The Terminal, Kubernetes, and Data engines can be installed now; Git and Chat engine builds are not published yet. Every app and every feature is free, with no tiers or paywalls, and one OpenEng account, created at auth.openeng.app, signs you in to all five.

Which operating systems do the five apps support?

Each app has its own install script. Engine builds for Terminal, Kubernetes, and Data are published for Linux and macOS; on Windows the engine runs through WSL. Git and Chat engine builds are not published yet. The browser side works in any modern browser. All five apps are in beta as of September 25, 2026.

Which operating systems does the AI engine support today?

A Linux x86_64 engine is live today. macOS and Windows builds will follow, and general availability of the platform is planned for Q4 2026.

Is OpenEng open source?

No. The OpenEng platform itself is not open source. Its engine runs open-weight (open-source) models such as Llama, Qwen, and Gemma entirely on your machine, and the company builds on and customizes open-source software such as Ollama, ONNX, and MCP for enterprises.

When is general availability planned?

General availability of the OpenEng AI platform is planned for Q4 2026, when the paid plans (Pro and Enterprise) will launch; the engine itself is free to run today. The five developer apps are a separate suite with their own launch, and they are in beta now.

Try the apps, or have us deploy the platform.

The five apps are free and in beta today. For a governed deployment of the AI platform on your own machines, talk to the services team.

Open openeng.app (opens in a new tab)Talk to us