offdataai
MCP server for AI agents

Give your AI agents the meaning behind your data

Connect Claude or any MCP client to offdataai. Agents get your approved concepts, metric definitions and join paths, ask your ontology questions in plain English, and run read-only SQL against your databases, all under scoped API keys.

  • From $1/month
  • 8 SQL targets
  • 174 templates
Claude · MCP · offdataaiConnected

tools/list

list_ontologies
get_query_context
search_concepts
get_concept
join_path
run_query
query_ontology
ask_ontology
search_catalog
propose_concept
Which customer segment drove revenue last quarter?
Any MCP client connects with an API key; answers follow the ontology's access rules.
Watch the demo

See it in action

A narrated walkthrough, recorded in the real app.

How it works

Three steps, minutes not months

1

Describe or import

Answer a short interview, open a template, draw on the canvas or connect a live database.

2

Review a validated model

Keys, grain, history and PII are checked before any code is written. Refine it in plain English.

3

Ship and keep it current

Native DDL, dbt, seeds, docs and an ontology. Push to Git, diff versions, generate migrations.

Overview

Why teams choose offdataai

app.offdataai.com
A plain-English question answered with the join path and the SQL that ran

A plain-English question answered with the join path and the SQL that ran

  1. 01

    AI agents are good at writing SQL and bad at knowing your business.

  2. 02

    Point one at a warehouse and it has to guess which of three revenue columns is the right one and how five tables join.

  3. 03

    The answers already exist in your data model and ontology; the MCP server hands them to the agent.

  4. 04

    The simplest path is ask_ontology: the agent passes the question in words, offdataai plans it from your ontology's metrics and fields, runs it under the same access rules and masking as any other question, and returns a one-sentence answer with the query, the SQL that ran and the rows.

  5. 05

    An agent that prefers to write SQL calls get_query_context for the concepts, the metric's definition and the resolved join path, then runs a read-only query through run_query.

  6. 06

    When it finds a gap, it can propose a new concept, which waits for a person to approve.

Real output

The code it writes, not a sketch

This is the shape of what offdataai generates here, straight from the validated model. Copy it, download it, or push it to Git.

  • Get the query context for a question
  • Ask in plain English
  • Search and browse the ontology
mcp · tools/calljson-rpc
1→ tools/call ask_ontology
2 { "ontology_id": "…",
3 "question": "revenue by customer segment" }
4
5← Answer: Enterprise is the largest segment by revenue.
6 -- planned as {"metrics": ["revenue"],
7 "dimensions": ["Customer.segment"]}
8 -- ran on warehouse (snowflake), 3 row(s)
9 SELECT …
10 Customer segment | Revenue
11 …
12
13# Or step by step, writing the SQL yourself:
14→ tools/call get_query_context
15 { "question": "revenue by customer segment" }
16← concepts, metric definition, tables, join path
17
18→ tools/call run_query
19 { "connection_id": "…", "sql": "SELECT …",
20 "max_rows": 200 }
Capabilities

What agents can do

  • Get the query context for a question

    get_query_context returns the concepts, metric definitions, tables and resolved join path for a natural-language question in one call.

  • Ask in plain English

    ask_ontology takes a question in words, plans it from your ontology's metrics and fields (never free SQL) and answers it under the same access rules, row filters and personal-data masking as any other agent question. One AI credit per question.

  • Search and browse the ontology

    search_concepts, get_concept, get_ontology and join_path let an agent explore your business vocabulary and how concepts connect.

  • Run read-only SQL safely

    run_query allows one SELECT or WITH statement, in a read-only transaction with a timeout and a row cap. Writes and stacked statements are refused.

  • Suggest, never overwrite

    propose_concept and propose_relation put an agent's ideas in the review queue. Your ontology only changes when a person accepts them.

  • Scoped API keys

    Grant ontology:read, ontology:write, ontology:query, query:read or pii:read per key. Database access and unmasked personal data are opt-in, and every query attempt is recorded.

  • Built on your real model

    The context comes from the same model that generates your DDL and dbt project, so what the agent reads matches what is deployed.

0

Validated templates

0

SQL targets

0

Modeling paradigms

0+

Industries covered

Generates native SQL and dbt for 8 platforms · connects live to Snowflake, Databricks, BigQuery, Redshift and Postgres

SnowflakeBigQueryDatabricksRedshiftPostgreSQLSynapseFabricDuckDBdbtSnowflakeBigQueryDatabricksRedshiftPostgreSQLSynapseFabricDuckDBdbt
FAQ

Frequently asked questions

What is an MCP server?
The Model Context Protocol (MCP) is an open standard that lets AI assistants such as Claude call external tools. offdataai's MCP server exposes your ontology and connected databases as tools, so an agent can look up what your data means before it writes SQL.
Why do AI agents need an ontology to query a warehouse?
Given only table names, an agent guesses: which table holds revenue, whether amount includes tax, how orders join to customers. The ontology answers those questions with approved definitions and resolved join paths, so the SQL an agent writes matches how your team actually calculates things.
Which tools does the offdataai MCP server provide?
list_ontologies, get_ontology, search_concepts, get_concept, get_query_context, join_path and search_catalog for reading the ontology; query_ontology, ask_ontology, get_record and get_related_records for answering questions through it; list_connections and run_query for querying connected databases; and list_proposals, propose_concept and propose_relation so agents can suggest additions for human review.
Can an agent change my data or my ontology?
No data can be changed. run_query accepts only a single SELECT or WITH statement, runs it in a read-only transaction with a statement timeout and a row cap of at most 1,000 rows, and records every attempt. Agents cannot edit the ontology directly either: propose_concept and propose_relation add suggestions to the review queue, where a person decides.
How is access controlled?
Every call is authenticated with an offdataai API key, and each key carries scopes: ontology:read for browsing, ontology:write for proposals, ontology:query for asking questions through the ontology, and query:read for running SQL. Questions asked through the ontology, whether structured or in plain English, follow its access rules and row filters, and personal-data fields stay masked unless the key also has pii:read. query:read and pii:read are opt-in and never granted to a key by default.
Which plan includes the MCP server?
Claude can design data models in the Model Builder from the Explorer plan up. The ontology tools, the public API and API keys come with Pro and above, and so do the live database connections that questions, run_query and importing a model from a database read from. Each plain-English question through ask_ontology uses one AI credit from your plan; structured questions through query_ontology use none. See the Claude connector docs at offdataai.com/docs/claude-connector for setup.
Which AI clients work with it?
Any client that speaks the Model Context Protocol over HTTP. In Claude, add it as a connector: you sign in to offdataai and approve what Claude may do, with no API key to copy. The server is stateless JSON-RPC, so it also works from your own agents and scripts with an API key.

Part of the offdataai platform

The business context behind accurate AI answers, built from the data model up.

See all six stages →
  1. 01DesignLogical, physical and dimensional models.
  2. 02GenerateSQL, dbt, documentation and migrations.
  3. 03DefineBusiness metrics, semantic models and ontology.
  4. 04ConnectReal warehouse schemas and metadata.
  5. 05GovernDrift, impact analysis, tests and approvals.
  6. 06ActivateMake trusted context available to AI agents.

This page covers Activate. Every stage works on the same model, so what you design here carries through to the rest.

Accuracy starts with context

Give your AI agents context they can trust.

Design the model, define what your metrics mean, and serve that context to every agent and BI tool, on the warehouse you already run.

Book a demoGet started

Self-serve plans available