One React codebase, six platforms, one missing backend

What it does

ibl.ai/os is the open-source frontend for the ibl.ai platform: a React 19 interface for building, configuring, and chatting with AI agents. It supports RAG document training, voice calls over WebRTC, multi-tenancy, SSO, and Stripe billing integrations. The same codebase wraps inside Tauri to produce macOS, Windows, Linux, iOS, and Android builds alongside the web and Chrome extension versions.

The interesting bit

Rather than maintaining separate native codebases, the project wraps one Next.js app in Tauri and ships it as web, desktop, and mobile. It is essentially a polished, skinnable enterprise chat client, despite the “OS” branding.

Key highlights

• Bring-your-own-LLM support across OpenAI, Anthropic, Google, Microsoft, Meta, and self-hosted models.
• Enterprise multi-tenancy with SSO via OAuth, OIDC, or SAML, plus RBAC and per-organization branding.
• RAG pipelines that ingest file uploads, Google Drive, OneDrive, Dropbox, and web crawls.
• MCP server connectors for tools like GitHub, Notion, and Slack.
• Embeddable iframe widgets, custom domains, and API keys for external integration.

Caveats

• The backend is not in the repo: authentication, agent APIs, and data services require ibl.ai’s proprietary platform, so fully standalone self-hosting is impossible without an enterprise license.
• The “native” apps are webview shells around the web frontend, which may dampen expectations of deep OS integration.
• The quickstart path assumes you already have a tenant key from the hosted service, making the repository closer to a client SDK than a standalone starter.

Verdict

Organizations already using ibl.ai’s backend get a polished, forkable frontend they can brand and ship everywhere. Developers looking for a complete, self-contained open-source AI agent stack should look elsewhere.