iblai/os
Enables organizations to create and deploy customizable AI agents with support for multiple LLM providers, real-time chat, voice, document training, analytics, user management, SSO authentication, and embeddable widgets. Built with React 19, Redux Toolkit, Radix UI, and Tailwind CSS.
ibl.ai/os – Open‑source AI Agent Platform
What it is – A self‑hostable, MIT‑licensed web app (built with Next.js/Tauri) that lets you create, run and manage conversational AI agents. It works on every major platform (web, Chrome extension, macOS, Windows, Linux, iOS, Android) from a single codebase and can talk to any LLM you prefer (OpenAI, Anthropic, Google, Meta, or your own model).
Why it matters – Companies and institutions often need an internal “ChatGPT‑like” service that they can control, brand, and bill for. ibl.ai/os gives you:
- Full ownership of code and data (no vendor lock‑in).
- Multi‑tenant SaaS features out of the box: SSO/OIDC/SAML, RBAC, Stripe billing, whitelabeling.
- Built‑in RAG pipelines (document upload, cloud‑drive connectors, web crawl) and live web‑search.
- Real‑time voice calls (WebRTC via LiveKit) and screen‑sharing inside a chat session.
- Easy deployment: Docker, Next.js standalone build, or native Tauri apps that are just web‑view shells.
Core features
| Category | Highlights |
|---|---|
| Agent building | Configurable LLM, system prompts, tool integrations, safety filters; collaborative “Projects” with shared context; versioned Canvas for rich documents |
| Data grounding | RAG ingestion (files, Google Drive, OneDrive, Dropbox, site crawl); live web search; multi‑step “Deep Research” reasoning |
| Interaction | Text chat, WebRTC voice calls, screen sharing, embed via iframe or API |
| Ops & scale | Analytics dashboards, tenant isolation, SSO/RBAC, Stripe billing, whitelabeling, API keys |
| Extensibility | MCP (Model Context Protocol) servers to add custom tools; connectors for GitHub, Notion, Slack, etc. |
Getting started
# clone and install dependencies (pnpm required)
git clone https://github.com/iblai/os.git
cd os
pnpm install
# quick dev run (you need an ibl.ai tenant key)
cp .env.example .env.local # edit NEXT_PUBLIC_MAIN_TENANT_KEY
pnpm dev # opens http://localhost:3000
For a one‑click setup with Claude Code, run the /setup command inside the app – it will fetch your tenant key and write .env.local for you.
Deployments
- Web –
pnpm build && node server-wrapper.jsor Docker (docker build -t os . && docker run -p 5000:5000 --env-file .env.local os). - Native apps – built with Tauri; set optional build‑time flags (
IBL_TENANT,IBL_ALLOW_IN_APP_PURCHASE) beforepnpm exec tauri buildto lock a tenant or enable in‑app purchases. - Enterprise – obtain a full‑backend license from ibl.ai and use the provided infra‑CLI for a complete SaaS deployment.
Testing – End‑to‑end Playwright suite under e2e/. Typical commands:
make e2e-install # install browsers (once)
make e2e # run all tests headless
make e2e-ui # interactive mode
Where to learn more
- Docs: https://ibl.ai/docs
- Development guide:
docs/development.md - Platform deployment guide:
docs/platform-deployment.md - Community / case studies: https://ibl.ai/case-studies
License – MIT (full source available).
Related
- Project
- Project
- Project
- Project