The ontology plane
for AI Agents.

Design semantic contracts in Studio. Publish the ontology to Hub. Deploy to Runtime — one click to a live MCP endpoint. Then run: any agent connects and queries your data by meaning, not by guessing column names.

Browse the Hub Start building free

For modelers

Design. Map. Ship.

Drag-and-drop classes and relationships. Map your database to the ontology with AI suggestions. Then publish the ontology to Hub to share it, or deploy to Runtime with one click for a live MCP endpoint on your own data.

Explore Studio →

For agent builders

Stable endpoint. No boilerplate.

Every Runtime deployment has a versioned MCP endpoint your agents connect to directly. Paste the URL into Claude Desktop or LangChain — your agent queries real data using semantic concepts, not guessed column names.

See how it works →

Design. Publish. Deploy. Run.

Studio produces two things: a portable ontology you publish to Hub, and a versioned OPS package (Ontology-Package Specification) — that same ontology plus your database mapping and an AI-ready tool schema — that deploys as a live cloud runtime. Agents connect via MCP with no infrastructure required.

1
Connect your database
Point at PostgreSQL, Snowflake, BigQuery and 7 more. We snapshot tables, columns, and keys.
2
Map schema to ontology
Auto-map with AI suggestions, or tune rules by hand — columns to classes and properties.
3
Publish to Hub
Push the ontology alone — no credentials, no mappings — to a versioned namespace.
4
Deploy to Runtime
One click ships your OPS package to OntoBoom Cloud. Stable MCP endpoint, no servers.
5
Run your agents
Paste the URL into Claude Desktop, Cursor, or LangChain. Agents query by meaning.

What's in the OPS package

Ontology-Package Specification — five files, one versioned zip.

ontology.ttl
OWL ontology in Turtle — classes, properties, relationships, constraints.
mapping.json
DB-to-ontology rules. Each rule maps a table or column to a semantic concept with confidence + reasoning.
tool-schema.json
AI-ready tool definition — drop into LangChain, OpenAI function calling, or any agent framework.
shapes.ttl + ops.json
SHACL validation shapes and a manifest with SHA-256 checksums for integrity verification.

Claude Desktop — claude_desktop_config.json

{
  "mcpServers": {
    "banking": {
      "url": "https://api.ontoboom.com/runtime/{id}/mcp"
    }
  }
}

LangGraph (Python) — same endpoint, any framework

from langchain_mcp_adapters.client import MultiServerMCPClient

async with MultiServerMCPClient({
    "banking": {
        "url": "https://api.ontoboom.com/runtime/{id}/mcp",
        "transport": "streamable_http",
    }
}) as client:
    tools = await client.get_tools()
    agent = create_react_agent(llm, tools)
    await agent.ainvoke({
        "messages": [("user",
            "Show investment accounts opened in the last 30 days")]
    })