CoreModels
Log in Start building
Schema infrastructure for AI

Connect your schemas once. Every agent builds on what your data means.

Import any schema. Map it to the others. Export it to the format each system needs. Give your agents one MCP endpoint to do the same.

Start a 14-day free trial Paste a schema ⌘V

14-day free trial. Builder $49/mo after the trial. Cancel any time.

26
connectors and formats
8
export targets: JSON Schema, JSON-LD, SQL, Avro, OWL, LinkML, ShEx, RML
14
day free trial · no charge during the trial
15
concurrent requests per agent
Why every agent build

Agents need one governed model of your data

CoreModels is a governed schema and meaning layer for agents. One model holds types, fields, definitions, allowed values, and how sources map. Platform, data, and AI builders put it in every agent build.

Agents connect to a hosted model over MCP: search, read, validate, export by default; writes need an admin-enabled server so agents consume without owning it.

For Platform engineers, data architects, and AI builders

  1. Align the model Define types, fields, and source maps

    Without: Agents infer meaning from names and local docs

    With CoreModels: Import schemas; align fields; AI suggests, humans accept

  2. Build the agent Write the agent and its contracts

    Without: Each agent encodes its own private meaning

    With CoreModels: Search, read, and export definitions over MCP

  3. Test Check payloads and contracts before go-live

    Without: Mismatches only show up in production

    With CoreModels: Validate payloads against the governed model

  4. Run Agent queries and acts on live data

    Without: Queries and outputs drift as sources diverge

    With CoreModels: Agent reads governed definitions before queries or replies

  5. Change Update schemas, sources, or allowed values

    Without: Agents break silently on schema change

    With CoreModels: Propose changes; find what breaks; humans approve

CoreModels — one governed model
CoreModels is the governed layer every stage reads from; writes only when admins enable it.

One governed model

Import schemas into one graph of types, fields, definitions, and source maps.

Catch breaks early

See what a schema change breaks before agents hit production.

Search before they act

Agents search, read, validate, and export over MCP before they act.

Humans keep control

Changes are proposed; people approve; the model keeps the record.

Connects JSON Schema, dbt, Snowflake, BigQuery, Salesforce, OWL; MCP with Claude, ChatGPT, Cursor, or your own agent

Recipes

Start from a recipe for the system you already run.

75 recipes, for 15 of the connectors.

Connector
Role

dbt · ga

Give your agent the governed meaning of status

An LLM writing SQL against your warehouse has three bad options: guess from the column name, read the SQL, or ask a human. Give it a fourth.

 dbt · early-access

The PR your dbt CI passes

Delete an accepted_values test and CI goes green, because the test is gone. Nothing compares the PR to what you agreed.

 jsonschema

Blank JSON Schema project

Turn a JSON Schema document into a governed model.

 neo4j

GraphRAG Grounding for Agents

Stop agents guessing your graph schema: give them the governed model over MCP.

 salesforce

Salesforce object-model governance

It takes forty seconds to change what a field means. Make the meaning governed.

 snowflake

Warehouse contracts for Snowflake

Your Snowflake schema is not a contract. Generate one from the governed model.

All 75 recipes →

Try it here · runs in your browser

Paste a JSON Schema. See the model.

Nothing leaves this page. Your schema is kept in this browser, so it is still here when you come back.

input.schema.json
model
Save this model to a workspace Sign in with GitHub or Google.

Import from · export to

Functions

Small functions. Call them from the console, the API or an agent.

Each one does one job on your schema graph.

import()

Import

JSON Schema, dbt, Snowflake, BigQuery, Salesforce, OWL and more. One graph.

map()

Map

Align fields across two schemas. AI suggests; you accept or edit.

validate()

Validate

Check a payload against the model before it reaches a system.

export()

Export

Generate JSON Schema, JSON-LD, SQL, Avro, OWL, LinkML, ShEx or RML.

Three lanes

What teams build on it.

AI building

A model agents can build on

Align the model and Build the agent: import what exists, propose types, generate schema and docs.

Align the modelBuild the agent
› model a returns flow on our orders schema
 Structured-data operations

Find breaks before production

Test and Change: import dbt metadata, audit drift, find what breaks, validate.

TestChange
› what breaks if we rename customer_id?
 Grounding

Answers grounded in production meaning

Run: the agent reads governed definitions before queries or replies.

Run
› which table holds active subscriptions?
Get going

From sign-in to first export in minutes.

01

Sign in

GitHub or Google. Your workspace opens on a 14-day free trial.

02

Import a schema

Paste it, point at a URL, or connect dbt and the warehouse.

03

Point your agent at it

One line adds the MCP server to Claude, Cursor or your own agent.

claude mcp add coremodels https://go.coremodels.io/mcp
Pricing

Start with a 14-day free trial. Subscribe when it is daily work.

Every plan starts with a 14-day free trial. Plans add seats, scale and support.

Builder

$49/mo

1 builder seat · up to 3 projects

See plan

Team

$990/mo

10 builder seats included · $99/month per additional seat · up to 10 projects

See plan

Enterprise

$4,000+

SSO · SLA · private vaults · a named contact.

See plan

AI builds. Humans govern.
Schemas connect both.

Start building Talk to us