Agentic Ontology

One ontology over every system, for any agent

The knowledge layer a health system's AI agents work from: patients, encounters, claims and providers from the systems you already run, certified once and served over MCP from a server in your network. 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 — so you can deploy anywhere, from your own cloud to a fully air-gapped network. No need to choose build vs. buy — you get both.

1.6M+Platform Users
400+Organizations
Syracuse UniversitySUNYGoogleNVIDIAMicrosoftOpenAIKaplan

“ibl.ai does amazing work and are great people.”

Paul LeBlanc, former President and CEO, Southern New Hampshire University
Don’t take our word for it?See what our partners say →
Want to read the source?View the code on GitHub →
ibl.ai: an ontology over your own systems
On the ibl.ai cloudStart at $0, usage-based from thereFree tier with renewable monthly credits, no credit card. Move to your own infrastructure anytime, nothing to rebuild.Hosted deploymentSee pricing
On your infrastructureRun it in your own environmentYour cloud, on-premise or air-gapped, under your control. Backend and infrastructure under a perpetual license.Full code license

Prefer to build together? Our AI Transformation team builds agents and apps with yours, on your systems.

How is Agentic Ontology different from a managed knowledge layer?

A managed AI vendor's knowledge layer usually means copying your data into the vendor's cloud and renting access to a model of your own organization. Agentic Ontology runs on your server, reads your systems where they are, and the code is yours.

Where the data goesAgentic OntologyRead in place by a server in your network; the ontology, passwords and MCP servers stay there.A managed knowledge layerCopied into the vendor's cloud.
Who holds the codeAgentic OntologyYou do: MIT-licensed, built from source on your own server.A managed knowledge layerThe vendor; you rent access.
What an agent can runAgentic OntologyCertified tools with fixed definitions and fixed SQL; raw SQL only if you add that toolset.A managed knowledge layerWhatever the vendor's connector exposes.
Who sees whatAgentic OntologyOne ibl.ai policy per toolset, callers signed in by ibl.ai, Entra ID or your own OIDC issuer, every call logged.A managed knowledge layerThe vendor's accounts and service keys.
Which agentsAgentic Ontologyibl.ai agents, Claude, Codex, Cursor or any MCP client.A managed knowledge layerThe vendor's own assistant.
PricingAgentic OntologyThe ontology is MIT-licensed; the ibl.ai platform is usage or a flat license, with no per-seat pricing.A managed knowledge layerPer seat, per connector, per year.

What is ibl.ai's Agentic Ontology?

Agentic Ontology is ibl.ai's on-premise knowledge layer: it connects the databases and HTTP APIs your organization already runs, turns them into one set of business terms that your people certify, and publishes those terms as MCP servers any agent can call, from a server inside your network where you own all the code and the data.

It is open source under the MIT license and built from the public source on your own Ubuntu server. The ibl.ai platform supplies sign-in, roles, API keys and agents; your server keeps the ontology, the database passwords and the MCP servers.

An agent never gets arbitrary SQL unless you add the raw toolset: each certified term becomes a tool with a fixed definition and fixed SQL, and each caller sees only the toolsets its grant reaches.

MCP Architecture Guide for Healthcare

How to connect your healthcare systems (EHR, LMS, credentialing, patient portals) to AI agents — with HIPAA compliance.

Read the Guide

How does the ibl.ai ontology turn your systems into tools an agent can call?

Five steps, one page of the ibl.ai ontology console each. The screenshots come from the console's demo mode, which serves fixed sample data, so every name and address in them is made up. Select a screenshot to enlarge it.

  1. Step 1 · Systems

    Connect

    Add a SQL connection or an HTTP API connection. SQL connections use a read-only login, credentials are tested before they are saved, and the password is kept on your server and nowhere else.

    Step 1, Connect: the Systems page. Select to enlarge.
  2. Step 2 · Map

    Dissect

    The engine reads the tables, profiles the columns and names the business terms above them. The Map draws that semantic layer over the physical one, so you can see which tables each definition came from.

    Step 2, Dissect: the Map page. Select to enlarge.
  3. Step 3 · Ontology

    Model

    Review what was inferred and certify it: object types, their properties, links and metrics. Personal columns are flagged, and only the definitions you certify are compiled into tools.

    Step 3, Model: the Ontology page. Select to enlarge.
  4. Step 4 · MCP Gateway, New MCP Server

    Expose

    Pick the toolsets, how callers sign in, and the origins and hosts allowed to reach the server. The config file is shown exactly as it will be written to your server, before anything is deployed.

    Step 4, Expose: the MCP Gateway, New MCP Server page. Select to enlarge.
  5. Step 5 · MCP Gateway, the server

    Serve

    Deploy, attach the server to an ibl.ai agent with a key of its own, or paste a ready-made snippet into Claude, Codex or Cursor, then watch the calls arrive. Each server keeps its revisions, and Roll Back restores the last one.

    Step 5, Serve: the MCP Gateway, the server page. Select to enlarge.
NVIDIAGoogle

Built on NVIDIA and Google. NVIDIA's auto-ontology engine infers the terms and Google's MCP Toolbox serves the tools, both running on your own server beside the ibl.ai console.

How do you install the ibl.ai ontology and connect a database to an agent?

The ibl.ai ontology console's Map during the install walkthrough: a pipeline of completed steps beside the business terms found in a CRM's database

Install the Ontology on Your Own Server and Connect a Database to an Agent (12:39, YouTube)

A 12-minute walkthrough on the ibl.ai YouTube channel, from a fresh Ubuntu server to an ibl.ai agent answering questions about a CRM: the four-block install, connecting the CRM's database, the Map, publishing an MCP server and attaching it to the agent.

Read the install guide →

How does Agentic Ontology by ibl.ai work on the systems you already run?

Agentic Ontology by ibl.ai gives a health system one set of definitions over the systems it already runs: a provider is the same provider in credentialing, scheduling and the claims database.

It reads those databases and HTTP APIs in place, personal columns are flagged and their samples masked before any reaches a model, and your teams certify the terms it finds.

Open source under the MIT license, and yours, including full code access.

What problem does the ibl.ai ontology solve for AI agents?

Most agent integrations are wired one system at a time, and each one defines a customer, a case or a student its own way. The ibl.ai ontology reads the systems you already run into one set of terms, and serves those terms to every agent.
One ontology, every systemSQL connections and HTTP API connections become one set of business terms and relationships, so a record is the same record in every system that holds it.
People certify, agents callThe engine suggests terms, joins and metrics, and a person certifies them. A certified term becomes a tool with a fixed definition and fixed SQL, so an agent picks the right one and cannot widen it.
Questions in plain wordsWith a model key, the ask_question tool answers a question in natural language with certified SQL and returns the SQL it ran, so the answer can be checked.

What does the ibl.ai ontology model for healthcare systems?

The steps are the same everywhereConnect, dissect, certify, expose and serve. What changes is the vocabulary: the terms are named in the words your people already use.
TermsPatients, encounters, claims, providers.
SystemsConnected in place, by database connection or HTTP API: scheduling, credentialing, the claims database and the data warehouse.

Which systems can the ibl.ai ontology connect?

Dissected and served as toolsPostgreSQL, MySQL, Snowflake and ClickHouse, by connection URL, with a schema scope, a connection test and a read-only login.
Any system with an HTTP APINo connector to wait for: give a base URL and an OpenAPI document, or name the endpoints by hand, and each endpoint becomes a tool. Reads and writes go in separate toolsets, and the upstream credential stays on your server.
Dissected and mapped, not servedTrino and Databricks are read into terms and shown on the Map, but not served as tools.
SQL connections coming soonDuckDB, HeavyDB, Kyuubi, SQLite, SQL Server, Oracle and BigQuery. Until then, a system with an HTTP API can be connected that way today.

Which AI agents can call the ibl.ai ontology's MCP servers?

ibl.ai agentsAttaching a server to an ibl.ai agent is one dialog: the agent gets a key of its own, limited to the toolsets you pick, and its MCP tool is turned on.
Claude, Codex, Cursor and your own codeEach server has ready-made configuration for Claude Desktop and Claude Code, Codex CLI, Cursor and Windsurf, the Python MCP SDK and cURL.
Browser-based agentsThey work once their origin is on the server's Allowed Origins, which the gateway checks before the credential.
Rate limits120 requests or 30 tool calls a minute, per caller.

How does the ibl.ai ontology control what each person and agent can see?

Read-only SQLSQL connections use a read-only login, so no agent can change your data through them. With an HTTP API you may allow writes, in a toolset of their own that an agent reaches only when a grant names it.
One policy per toolsetWho may call what is decided on the ibl.ai platform: one role and policy per toolset, granted to groups, users and agents. Two callers can get different tool lists from the same address.
Sign-in, per serverCallers sign in with ibl.ai credentials, Microsoft Entra ID or a custom OIDC issuer, chosen for each MCP server.
Down to the rowLimit what a caller sees by the toolsets a grant names, by the certified SQL a caller cannot widen, or by binding a parameter to a claim of the caller's verified sign-in.
Personal dataColumns that look personal are tagged and flagged in the Ontology, their samples are masked before any reaches a model, and you decide per toolset which tools may return them.
A log of every callEvery tool call and every refusal is logged on your server and shown under Activity, and the log can be shipped to your SIEM. Access is taken away by revoking a key or removing a group from a policy.

How do you install and run the ibl.ai ontology on your own server?

What the server needsUbuntu 22.04, 24.04 or 26.04 on x86-64 with 4 GB of memory, 15 GB free, a host name pointing at it and ports 80 and 443 open, plus an ibl.ai platform and an administrator of it.
Four blocks to pasteName the platform and the host, install Docker, build the console from source, start everything. Only the first block needs your values, and blocks two to four can be pasted again if one fails.
Built from source on your serverThe console is built from the public repository and run by systemd; nothing of ibl.ai's arrives as a prebuilt image. The ontology engine, its Postgres and Caddy, which gets the HTTPS certificate, run beside it in Docker.
What is exposedOnly ports 80 and 443, through Caddy, and anything behind them that returns data asks for a sign-in or a key. The engine, its database and each MCP server's process listen on the server's loopback only.
A model is optionalConnecting systems, the Map, certifying terms and the tools all work without one. An OpenAI API key, or any gateway that speaks OpenAI's API, turns on inferred joins and questions in plain words.
Looking after itSettings, upgrades, backups and what to do when something stops are in the operations guide. Each MCP server keeps its revisions, and Roll Back puts the previous one back.

Why run the ibl.ai ontology instead of a managed vendor's knowledge layer?

Where the data goesAgentic Ontology: Read in place by a server in your network | Managed vendor: Copied into the vendor's cloud
Who holds the codeAgentic Ontology: You do, MIT-licensed and built from source | Managed vendor: The vendor; you rent access
What an agent can runAgentic Ontology: Certified tools with fixed SQL | Managed vendor: Whatever the vendor's connector exposes
Which agentsAgentic Ontology: ibl.ai agents, Claude, Codex, Cursor or any MCP client | Managed vendor: The vendor's own assistant
PricingAgentic Ontology: MIT; the ibl.ai platform is usage or a flat license, no per-seat pricing | Managed vendor: Per seat, per connector, per year
ExitAgentic Ontology: The server, the ontology and the code stay with you | Managed vendor: The model of your organization leaves with the vendor

How does the ontology fit into the rest of the ibl.ai platform?

The ibl.ai platform supplies sign-in, roles, API keys and agents; Agentic OS runs the agents that call the ontology's MCP servers. One sign-in and one set of roles across the platform, and you own all the code and the data.
Installed with our teamIn an AI Transformation engagement, ibl.ai's engineers install the ontology with your team, connect your first systems and certify the first terms, then stay on as your partner as new systems, agents and updates arrive.

How can you try the ibl.ai ontology today?

Watch the walkthroughA 12-minute video on the ibl.ai YouTube channel, from a fresh Ubuntu server to an agent answering questions about a CRM.
Run it in demo modeClone github.com/iblai/ontology and run it with NEXT_PUBLIC_DEMO=1: sample data, no ibl.ai account and no engine.
Install it on your serverFollow the install guide at ibl.ai/developer/guides/ontology-install: four blocks to paste on a fresh Ubuntu server.
Talk to our teamBook a demo, or contact ibl.ai to plan an install on your own infrastructure with our engineers.
  • Free

    Get started

    $0per
    month

    No credit card required

    Get Started Free

    This includes:

    • Monthly credits to try agents, renewed each month
    • Build agents on any model, trained on your own documents
    • Chat on web, desktop and mobile
    • Documentation and community support

    For one person trying things out

  • Most popular

    Pro

    For everyday work

    $15per
    month

    Billed monthly

    Subscribe to Pro

    Everything in Free, plus:

    • More credits every month, with auto-recharge when you want it
    • Voice, workflows and your tools through MCP
    • Analytics, embedding and the API
    • Email support

    For professionals and small teams running agents every day

  • Max

    For all-day work

    $50per
    month

    Billed monthly

    Subscribe to Max

    Everything in Pro, plus:

    • A much larger monthly credit allowance
    • Evaluations, spend limits and monetization
    • Roles and teams for access control
    • Priority support and early access

    For teams and departments that depend on agents

  • Ultra

    For your largest deployments

    $100per
    month

    Billed monthly

    Subscribe to Ultra

    Everything in Max, plus:

    • The largest monthly credit allowance
    • Several organizations and single sign-on
    • Hand-off to human agents
    • Dedicated support and hands-on onboarding

    For organizations formalizing agents across many teams

  • Recommended

    Enterprise

    Own the Platform

    Custom quote

    Pilots from $15K · one-time

    Contact us

    This includes:

    • Full codebase ownership under a perpetual license
    • Deploy anywhere — on-prem, your VPC, air-gapped, or managed SaaS
    • Start with a pilot — prove value on real data first
    • Custom AI engineering — integrations, SSO, SIS, LMS, agents
    • Run any LLM — Claude, GPT, Gemini, Llama; switch anytime
    • SOC 2 & compliance — enterprise security with audit support

Enterprise — Own the Platform. Own the entire platform. We transfer the full source under a perpetual license and deploy it on-premise, in your cloud, or air-gapped — zero vendor lock-in, run any LLM. Start with a pilot, then scale with our engineers wiring it into your data and systems.

Self-serve prices in USD, billed monthly; change or cancel your plan anytime from Billing in the app. Enterprise is scoped with our team: a pilot from $15K, a one-time deployment, or full codebase ownership on a custom quote.

What our partners say about us

Chris Gabriel

Chris Gabriel | Google

Lorena Barba

Lorena Barba | George Washington University

Dr. Juana Mendenhall

Dr. Juana Mendenhall | Morehouse College

Juile Diop

Juile Diop | MIT

Adam Tetelman

Adam Tetelman | Nvidia

Jason Dom

Jason Dom | American Public University System

Benjamin Breyer

Benjamin Breyer | Columbia University

Ken Fujiuchi

Ken Fujiuchi | SUNY

Erika Digirolamo

Erika Digirolamo | Monroe College

David Flaten

David Flaten | SUNY

David Vise

David Vise | Modern States Education Alliance

Linda Wood

Linda Wood | ARM Institute (U.S. Department of Defense)

Chris Gabriel

Chris Gabriel | Google

Lorena Barba

Lorena Barba | George Washington University

Dr. Juana Mendenhall

Dr. Juana Mendenhall | Morehouse College

Juile Diop

Juile Diop | MIT

Adam Tetelman

Adam Tetelman | Nvidia

Jason Dom

Jason Dom | American Public University System

Benjamin Breyer

Benjamin Breyer | Columbia University

Ken Fujiuchi

Ken Fujiuchi | SUNY

Erika Digirolamo

Erika Digirolamo | Monroe College

David Flaten

David Flaten | SUNY

David Vise

David Vise | Modern States Education Alliance

Linda Wood

Linda Wood | ARM Institute (U.S. Department of Defense)

Frequently Asked Questions

Bring us a hard problem. Keep the code.

Run Agentic Ontology on your own terms

Start free on our cloud today, or take the whole codebase into your own infrastructure. Either way our engineers work alongside your team, and you own all the code and the data.

ibl.ai is family-owned and operated from New York, NY — a U.S.-headquartered, domestically-owned long-term partner, not a vendor that sells licenses and moves on. · SOC 2 Type II certified