A reusable runtime for themeable real-time AI command interfaces on top of the real OpenClaw Gateway.
Not just another chat skin.
A runtime for battle-ready AI terminals.
OpenClaw Hub Runtime is a modular frontend runtime for AI command surfaces with streaming UX, reconnect-aware diagnostics, resume/recovery foundations, and reusable shell bindings.
- Real backend, not a mock demo — built around actual OpenClaw Gateway integration
- Runtime, not a one-off shell — one runtime already powers both a flagship Jarvis shell and a minimal shell
- Resilience is part of the product — reconnect, recovery, diagnostics, and resume flows are first-class
- Designed to be reused — shell identity and runtime behavior are intentionally separated
See examples/minimal-hub/README.md and examples/minimal-hub/index.html.
- a flagship J.A.R.V.I.S.-style shell
- a reusable runtime core
- a real OpenClaw Gateway adapter
- a reconnect and recovery layer
- a deterministic test harness
- a minimal shell example proving portability
Build one runtime once, then launch multiple AI shells on top of it.
- OpenClaw Gateway running locally on
127.0.0.1:18789 - a valid local gateway token stored outside published history
- a local static file server
cd d:\openclaw
python -m http.server 8787Jarvis shell
http://127.0.0.1:8787/apps/jarvis-hub/index.html
Minimal shell
http://127.0.0.1:8787/examples/minimal-hub/index.html
Resume showcase
http://127.0.0.1:8787/apps/jarvis-hub/index.html?testResume=1
Poster showcase
http://127.0.0.1:8787/apps/jarvis-hub/index.html?testResume=1&poster=1
For a fuller setup guide, see docs/getting-started.md.
- real OpenClaw Gateway WebSocket connection
connect,health,chat.history,chat.send,agents.list- soft-throttled streaming assistant output
- fenced code block rendering as sub-terminals
- sidebar agent list
- diagnostics-first reconnect UX
- resume / reconcile / replay foundations
- built-in
/diagnoseand/test-resumeshowcase commands - shared Chinese copy support
- reusable shell binding through
HUB_CONFIG
Star this if you care about any of these directions:
- reusable AI command interface runtimes
- real-time streaming UI on top of actual agent backends
- reconnect / recovery UX as a product feature
- shell portability instead of single-demo UI work
- OpenClaw-native frontend experiments with real integration
If you only want a pretty screenshot, this repo may be overbuilt. If you want a strong reference point for runtime + shell + resilience, that is exactly what this project is trying to become.
OpenClaw Hub Runtime is organized as layered modules:
- Hub Shell — layout, shell DOM, copy loading, theme styling
- Runtime — shell binding, store, renderer, commands, bootstrap
- Adapter — OpenClaw transport and protocol handling
- Reconnect Manager — resilience, replay, recovery
- Test Harness — deterministic showcase and resume scenarios
Read the full system breakdown in docs/architecture.md.
- Getting Started:
docs/getting-started.md - Architecture:
docs/architecture.md - Theming:
docs/theming.md - Adapter API:
docs/adapter-api.md - Launch Guide:
docs/github-launch.md - Launch Content Pack:
docs/launch-content-pack.md - Launch Kit:
docs/launch-kit.md - GitHub Publish Pack:
docs/github-publish-pack.md - Final Publish Preset:
docs/final-publish-preset.md - Star Pack:
docs/star-pack.md - Changelog:
CHANGELOG.md
apps/jarvis-hub/index.html
This is the cinematic HUD reference implementation.
examples/minimal-hub/index.html
This proves the runtime can boot a second shell with a simpler surface and different product feel.
apps/
jarvis-hub/
examples/
minimal-hub/
packages/
adapters/openclaw/
runtime/
shared/
test-harness/
themes/
docs/
scripts/
This repository is currently best understood as:
the first public reference runtime for building OpenClaw-native AI command interfaces.
It is already a strong working implementation, but it is also the beginning of a broader reusable runtime direction.
Contributions are welcome. Start with:
CONTRIBUTING.mddocs/architecture.mddocs/theming.mddocs/adapter-api.md
Please run the PowerShell verification scripts before proposing changes.
If you discover a sensitive issue, see SECURITY.md.
This repository is licensed under the MIT License. See LICENSE.

