Launching today
Agentic Operating System
Open-source AI agent platform — own it, run any model
14 followers
Open-source AI agent platform — own it, run any model
14 followers
ibl.ai OS is the open-source platform for building AI agents you own and self-host — your infrastructure, your data, your choice of model. Go from prototype to production in minutes: • RAG training + web-search grounding • Real-time voice (WebRTC), canvas & artifacts • MCP, multi-tenancy, SSO, RBAC • Web, desktop & mobile apps — MIT licensed • LLM-agnostic: GPT, Claude, Gemini, Llama • Drive it all from Claude via agent skills: github.com/iblai/api Fork it. Self-host it. Own it.
Hey hunters 👋 Miguel here — ibl.ai, family-owned out of NYC.
We kept hearing the same thing from universities, enterprises, and government teams: they loved what AI agents could do but hated renting them — $20–60 per seat, locked to one vendor's model, their data living on someone else's cloud. So we built the opposite and open-sourced it.
ibl.ai OS lets you build, deploy, and manage AI agents from prototype to production in minutes — RAG/document training, real-time voice (WebRTC), web-search grounding, canvas & artifacts, MCP integration, multi-tenancy, SSO, and RBAC — shipping as web, desktop, and mobile apps. Run any model (GPT, Claude, Gemini, Llama) and switch anytime.
Straight talk on the "open" part, since someone will ask: the repo is MIT and it's the full agent platform — the client and the apps. Fork it, white-label it, build on it. The backend (model routing, auth, data services) is what we run as a managed service so you can be live in minutes — and for teams that want everything, our enterprise tier hands over the entire stack under a perpetual license to self-host in your own VPC. We'd rather show you exactly where the line is than wave "open source" around.
160+ agent templates, and the platform already runs for 1.6M+ users across 400+ organizations.
And for builders: you can drive the entire platform straight from your AI coding agent. Run "npx skills add iblai/api" and you get 30+ agent skills mapped 1:1 to our REST API — spin up and configure agents, train datasets, manage users and RBAC, pull analytics — plus a hosted chat MCP server (iblai-agent-chat) to talk to deployed agents, tool use and RAG included. Works in Claude Code, Cursor, and Copilot — point your agent at the repo and it knows the whole API. → github.com/iblai/api
⭐ github.com/iblai/os → ibl.ai/os
What's the first agent you'd spin up — a tutor, an internal help desk, something else? Genuinely keen for the feedback. 🙏
Being able to swap between GPT, Claude, and Llama without rebuilding anything is genuinely refreshing, and the open-source angle makes it feel like mine to actually experiment with.