Feature

Custom Connectors

No connector for your source? Have one today.

QuickBooks had no connector. 25 minutes later it had one: 26 endpoints loading into BigQuery. A dashboard and a data model now sit on top. Your AI assistant writes the connector from the API docs. You test it on live data; OWOX keeps every version and runs it on schedule into your own data warehouse. Documentation

Get started free

Custom Connectors

Why OWOX

When a report needs a source OWOX has no connector for, the usual answers are a vendor roadmap, a script only its author can fix, or a CSV exported by hand every month. With Custom Connectors your AI assistant writes that connector from the service's docs, and OWOX runs it like a built-in one.

  • QuickBooks went from no connector to a 26-endpoint connector in BigQuery in 25 minutes; a dashboard and a data model now run on its data.
  • OAuth2 with token rotation, retries, backfills and schedules are built in.
  • Every publish is a version you can roll back, and the data lands in your own data warehouse.

Hand this to your AI

The source your report needs has no connector, so its numbers arrive as a monthly CSV pasted in by hand.

Paste this into your AI assistant. It writes a custom connector from the service's API docs; you test it on live data, and OWOX runs it on schedule into your data warehouse.

The prompt

write an OWOX custom connector for the service I name from its API docs, one Data Mart per endpoint, then help me test it on live data and schedule a daily load into my data warehouse

Trust

The Data Partner You Can Trust

users

166K Source: Internal OWOX data as of 2026-11-15

Executives and data analysts across 20+ industries

G2 rating

4.9★2

“A trusted data analytics partner” — verified review

npm installs

171K3

In Open Source We Trust

Customers

Docs In, Connector Out

The source your report is missing becomes a connector you own

Every team has one: the affiliate network, the accounting system, the niche ad platform whose numbers arrive as a monthly CSV because nobody will build the connector. A custom connector ends the wait. Your AI assistant writes it from the service's API docs, OWOX runs it on live data, and from then on the source feeds a Data Mart like any built-in source. QuickBooks was built this way: OAuth2, nested financial reports, 26 endpoints.

THE ANALYST

From API docs to a running connector

Give your assistant two documents: the service's API reference and OWOX's connector reference. It returns the connector as a JSON manifest. Paste it in or import the file, run it on live data, read the rows, publish. OWOX proposes the fields and their types.

The assistant drafts. OWOX runs it in production.

THE BUSINESS USER

The cost line that was a monthly CSV lands next to every other source

Once a Data Mart sits on the connector, its data refreshes on schedule alongside Google Ads, Shopify and everything else you already load. The number that used to be pasted from an export is now a column in your data warehouse, with history.

No more hand-exported CSV.

Not a script only its author can fix — A custom connector gets what a one-off script never does: versions you can roll back, test runs with logs, retries when the service is busy, token rotation handled, and a Data Mart that runs on a schedule.

For Data Teams

For Data Teams

Your assistant drafts, OWOX carries the hard part

The draft is the easy half. The half that breaks scripts at 3 a.m. is built into the runtime.

OAuth2 with token rotation, handled

Refresh-token grants, rotated tokens persisted, retries on failure, and the rate limits you set respected. QuickBooks rotates its refresh token on every use, and the connector saves each new one.

Every page, every date, every account

Long result lists, incremental loads, backfills for past dates, several accounts in one run, and reports the service prepares in the background.

Your keys and data never reach the assistant

The assistant writes the connector, nothing more. Keys and tokens are Secret parameters entered on the Data Mart page, never stored in the connector.

For Business Teams

For Business Teams

One connector, every endpoint, each its own table

A source is never one table. Each endpoint feeds its own Data Mart with fields, a destination table and a schedule, so the whole system lands, not a sample of it.

Paid traffic with its cost next to it

When an affiliate network or a niche ad platform sends sessions and orders but has no connector, its spend loads into the same data warehouse as the traffic it drove.

The books joined to the orders

Invoices, bills, payments and the P&L from an accounting system join revenue and ad spend, so finance and marketing work from the same figures.

AI and SaaS spend beside every other cost

Usage bills from AI and SaaS vendors with an API land in the same cost table finance already reads.

How it works

How a custom connector comes together

Step 1: Give your assistant two docs

The API reference of the service you need, and OWOX's connector reference.

Step 2: Bring in the connector

Paste in what your assistant writes, or import it as a file.

Step 3: Test on live data

Run it, read the rows and the logs, let OWOX propose the fields and their types, fix what the draft missed.

Step 4: Publish and build the Data Mart

Publishing saves a version. Pick the connector in Data Mart setup, enter keys and tokens there, set a schedule. Done.

Editions

Available in all editions

Pay as you go

Top up only when you have a decision to make. No subscription loop.

$0.30 / run

Start free

Team

Your data partner for your leadership team.

$790 / month

Book a demo

Scale

Your data partner for the whole company.

Let's talk

Book a demo

See full pricing & comparison →

Why teams pick OWOX

Everything business needs

Proven on a real source

QuickBooks: 26 endpoints described field by field, an accounting dashboard on the loaded data, and a step-by-step guide to reproduce it. The connector went from first draft to production in 25 minutes.

Versioned, with rollback

Every publish creates a version. A draft leaves the running one untouched, and an earlier version can be made active again.

Secrets stay in OWOX

Keys and tokens are Secret parameters, entered on the Data Mart page and never stored in the connector or shown to the assistant.

Lands in your own data warehouse

The data loads into your own data warehouse, BigQuery, Snowflake, Redshift, Athena or Databricks, as a Data Mart with fields, a destination table and a schedule, ready for SQL, dashboards and AI assistants.

FAQ

Frequently asked questions

Can I connect a missing source myself?

Yes. If the service has an HTTPS API that answers in JSON, CSV or JSONL and accepts an API key, a token or an OAuth2 refresh token, give its documentation and OWOX's connector reference to your AI assistant, and it writes the connector. Data Owners and Project Admins publish it, and anyone who sets up a Data Mart can use it. The connector reference lists exactly what a connector can do.

Has a real source been built this way?

Yes. QuickBooks: 26 endpoints described field by field, an accounting dashboard on the loaded data, and a step-by-step guide to reproduce it in your own project. The connector went from first draft to production in 25 minutes.

Do I need to write code?

No. Your assistant writes the connector. You say which data you want, check the test run and publish. A form in OWOX walks you through the settings if you would rather edit by hand.

Where does a custom connector run?

In OWOX, like every built-in connector. It runs on the Data Mart's schedule or on demand, and loads the rows into your own data warehouse.

Will a custom connector load the same way every time?

It runs the version you published until you publish another. A Data Mart follows the active version or stays pinned to an earlier one, and a draft never touches the running version. If the service fails or says it is busy, the request is retried, and the connector can be told to send no more requests than the service allows.

How do I keep a custom connector working when the service changes?

Export the connector, give the file and the service's updated docs to your assistant, and ask it to keep the existing field names, because your Data Marts use them. Test it and publish it as a new version. If anything breaks, make the earlier version active again. The connector is yours, and so is every version of it.

Is it enough to give my AI assistant the docs?

For a first draft, yes. For publishing, run the test: a draft can miss a reserved field name or a pagination rule, and the logs show it in one run.

What users are saying

Every claim has a receipt.

Real things real customers said — each quote pinned to a specific claim, straight from the quotes database.

A3re: trusting AI
Nodari RizunNodari RizunFounder & CEO, Pürblack®
“AI, by the nature of the models, will hallucinate. And because of that, you need something which will create guardrails to ensure that there are no hallucinations, that you can trust your data.”
C9re: one source of truth
Mark SimmonsMark SimmonsCMO, Pürblack®
“We had six or seven different channels and no single source of truth — it was almost impossible, a lot of guesswork.”
E7re: getting time back
Nodari RizunNodari RizunFounder & CEO, Pürblack®
“We regained time. And time is the one resource that never comes back.”

Recent updates

Missing a source? Hand your assistant the docs.

Test the connector it writes on live data and publish it. From the next scheduled run the source loads into your own data warehouse, as if it had always been there.

Get started free