For Developers

Build on an AI platform you own

ibl.ai is the agentic AI platform where you own all the code and the data. Its 5 repositories are public on GitHub β€” the application layer, the SDK, 161 agent skills and two MCP servers, and the infrastructure CLI β€” so you can read the source, scaffold an app, and run the whole stack on your own cloud, on-premise, or fully air-gapped, model-agnostic across any LLM and with no per-seat pricing.

Not an API you rent β€” a full platform you keep. 1.6M+ users across 400+ organizations run the platform this way, including NVIDIA, MIT, and Syracuse University.

161 skills Β· iblai/base as of (b7569f8)

Three doors

Where do I start on ibl.ai?

Developers come to ibl.ai to do one of three things. Each door below is one command and the three pages that take you from there to running code β€” every one generated from the repository it documents, so it cannot drift.

Build an agent

Create an agent, give it a prompt, tools, memory and a dataset, test it against a benchmark, and talk to it from any MCP client β€” with a screen or headless over REST.

$ npx skills add iblai/base --all

from github.com/iblai/base

Build an app

Scaffold a production Next.js app with SSO, agent chat, profiles, notifications and analytics wired to the platform, then add features one skill at a time.

$ /iblai-vibe-start

from github.com/iblai/base

Run the platform yourself

Provision the whole stack on your own AWS or GCP account with one interactive CLI, or bootstrap an existing server β€” the CLI and its docs live in github.com/iblai/infra-cli.

$ pip install ibl-cli && ibl setup

from github.com/iblai/infra-cli

161 skills, one install

161 skills Β· iblai/base as of (b7569f8)

Which skill builds the part of the agent I need?

Every skill is a how-to a coding agent runs and a human can read. Search by what you are trying to do, or start from the parts of an agent below β€” each has a skill with a screen (iblai-vibe-*) and a headless twin over REST (iblai-api-*).

All 161 skills by folder and kind Β· What a skill kind means

Written for agents

How do I set up my coding agent for ibl.ai in one step?

  1. 1. Install every ibl.ai skill for Claude Code, Cursor and Codex β€” the same command refreshes them later.

    $ npx skills add iblai/base --all
  2. 2. Type the first skill in your coding agent; it asks what you are building, connects your organization and writes your credentials to .env.

    $ /iblai-vibe-start
  3. 3. Give the assistant the SDK's components, hooks and page templates while it writes your app. Details

    $ npx -y @iblai/mcp
  4. 4. Point any MCP client at a deployed agent for a live conversation β€” streamed responses, tools, retrieval. Details

  5. 5. Everything on this site is readable by an agent: llms.txt, a Markdown twin of every page, and the agent-skills discovery index. Details

Highlighted repositories

Which ibl.ai repositories can I read and run myself?

These 5 highlighted repositories are the heart of the platform β€” not every public repo; github.com/iblai lists them all. Most AI vendors give you an endpoint; ibl.ai gives you the source you can read, fork, run, and extend. Every card below shows the live release straight from GitHub.

Application layer
iblai/os

The open-source platform itself. Create and deploy customizable AI agents with multi-LLM support, real-time chat and voice, document training, analytics, user management, SSO, and embeddable widgets.

v0.202.2released Β·updated Β·TypeScript 639Powers Agentic OS

Scaffold a production-ready Next.js app on ibl.ai β€” SSO auth, AI chat, profiles, notifications, and analytics wired up out of the box, with pre-built components and Claude Code skills.

v2.24.0released Β·updated Β·TypeScript 234Powers Agentic Base
$ npx skills add iblai/base

An interactive CLI that provisions the whole platform on your own AWS or GCP account with Terraform and configures servers with Ansible β€” and it can bootstrap any existing server.

v2.0.0released Β·updated Β·Python 12
Reference
iblai/docs

The documentation source β€” every page published at ibl.ai/docs, versioned in the open and available for contributions.

v4.227.15released Β·updated Β·Markdown 9
Application layer
iblai/lms

Skills-intelligence platform on the same SDK β€” discover courses, track competencies, and earn credentials. A second complete application you can read end to end.

v0.104.2released Β·updated Β·TypeScript 96Powers Agentic LMS
How-to, by goal

How do I deploy and run ibl.ai on my own infrastructure?

Cloud, VPC, on-premise or fully air-gapped β€” each answer below is a page generated from the repository that ships the code, so the steps are the ones the engineers run.

Compare

How does ibl.ai compare with other developer platforms?

The one axis where every hosted platform is identical and ibl.ai is the opposite: you receive the source code and keep the data. Each comparison below leads with that and then goes feature by feature.

Shipping velocity

What changed in ibl.ai recently?

The platform ships a changelog roughly every week, grouped by service image. The three most recent:

  • A Classifier Step That Routes Each Request to the Right Agent

    Agent workflows on ibl.ai gain a Classifier step: a decision model sorts each message into 2 to 20 categories, a yes or no, or a scale, sends it down that path, takes a Fallback below a confidence threshold (0.6 by default), and can be tried on a sample message before you publish.

  • Platform Update β€” November 13, 2026

    Eleven services and packages moved in one week, led by a workflow classifier node that routes runs by model decision, an agent-chat MCP endpoint that runs a full agent turn as a Claude Desktop or claude.ai connector, agent channel identity linking with a Microsoft Teams connector, a credits-only billing flow with a plan picker, RBAC on the Create Agent form, a privacy audit log with PII and PHI filtering, and cross-organization agent copy.

  • User Spend Limits: One AI Budget per Person, Every Agent

    A fourth Billing tab, User Limits, caps what one person spends across every agent in a workspace, and a Default for New Users gives each member who joins later their own limit automatically, with block or alert-only enforcement and near-limit alerts.

All platform updates Β· RSS

A store of value, not a subscription

What do I actually own when I build on ibl.ai?

Rented AI is a cost that recurs forever and leaves nothing behind. An owned platform is an asset: the code, the data, and the infrastructure are yours, and they keep their value after the contract ends.

Rented AI

  • Per-seat or per-token billing that scales with headcount
  • Locked to one vendor's models and roadmap
  • Your data lives in someone else's cloud
  • Credits evaporate when the subscription lapses

ibl.ai β€” owned

  • Full source under a perpetual license
  • Model-agnostic β€” run any LLM, switch anytime
  • Code, data, and infra stay in your environment
  • Usage-based, never per-seat β€” yours to keep

1.6M+ users across 400+ organizations run the platform this way, including NVIDIA, MIT, and Syracuse University. ibl.ai is family-owned and operated from New York, NY. Explore the self-hosted platform or the full-code license.

Clone a repo and run it today

On ibl.ai you own all the code and the data, run it model-agnostic across any LLM, and pay with no per-seat pricing β€” so you can deploy anywhere, from your own cloud to a fully air-gapped network.

Developer FAQ