Skip to main content

​
Loong Documentation

Loong is a Rust-built base for vertical AI agents. It starts from a runnable local assistant path, then grows outward through explicit runtime ownership, governed capabilities, and extension seams that teams can build on without rewriting the core. This site stays short at the top, then branches into focused tracks instead of turning the homepage into a giant matrix.

​
Choose A Track

If you want to…Start hereWhat you get
install Loong and reach first value fastGet Startedinstall, onboard, first-run, and doctor guidance
follow one complete provider-plus-channel rollout pathCommon Setupsend-to-end playbooks for hosted, service-channel, local, and gateway-owned setups
understand the shared public config shape firstConfiguration Patternsprovider profiles, channel accounts, gateway selectors, memory, and trust toggles
unblock managed skills, browser preview, or web policy gatesManaged Skills and Browser And Web Boundariesinstall flow, persisted policy, preview enablement, and private-host/domain troubleshooting
configure a provider or channel right nowProvider Guides and Channel Guidesone page per built-in provider or channel, with recipes kept for representative rollout patterns
understand the public runtime model before configuring itUse Loongthe current surface shape, page map, and capability boundaries
extend or evaluate the architectureBuild On Loongcrate boundaries, contribution workflow, and builder-facing contracts
understand why the project exists and what product stance shapes itWhy Loongthe public rationale, positioning, and product posture in one page
check public commitments and project policyReferenceroadmap, docs policy, reliability, security, and releases

​
Public Docs Layers

The public docs are intentionally split into three layers:
  • README.md stays as the repository landing page.
  • site/ is the main reader-facing docs surface that Mintlify deploys.
  • docs/ stays repository-native for source-facing specs and supporting references.
That split keeps the front page short without hiding the deeper public contract from source readers.

​
Known Target Stack?

If the rollout shape is already clear, skip the broad concept pass and go straight to a dedicated playbook.
If you already know you want…Start here
Volcengine plus Feishu or LarkVolcengine Plus Feishu Or Lark
Volcengine plus WeCom as the enterprise laneWeCom Rollout
BytePlus Coding plus TelegramBytePlus Coding Plus Telegram
a local or self-hosted model plus outbound deliveryLocal Model Plus Outbound Delivery
several runtime-backed channels on one hostGateway Rollout

​
What This Site Optimizes For

  • A short first-run path instead of a giant landing page checklist.
  • Focused provider and channel pages instead of stuffing matrices into the README.
  • Concept pages for the runtime model and tutorial pages for concrete setup.
  • Public architecture and contribution material for builders.
  • Public reference material for roadmap, reliability, security, and releases.

​
Reading Rules

  • Start with Get Started if you need first value fast.
  • Jump to Common Setups when you already know the target rollout shape and want one end-to-end path.
  • Jump to Provider Guides or Channel Guides when you already know what you want to configure and need the exact built-in contract.
  • Keep Provider Recipes and Channel Recipes for representative rollout patterns and smoke-test flows.
  • Use Use Loong when you want the public runtime model before touching config.
  • Use Build On Loong and Reference after the operator path is clear.
If you want the explicit boundary and publishing policy, see Documentation Policy.