Connectors
A connector pulls a REST API into a typed Postgres table on autario and keeps it fresh on a schedule. The account owner sets one up in the interface, because it involves a credential, and everything after that is ordinary data: the tables a connector produces are queryable through the same endpoints as any other dataset.
Pull data from any tool that has an API (2:34)
Many tools we use have an API. autario pulls that data, turns it into a table that refreshes on a schedule and makes it available to BI tools, data apps and your AI, shown with Google Search Console.
Watch on YouTube: Pull data from any tool that has an API
What exists
There is a preset library covering ad platforms, product and web analytics, search and SEO tools, commerce and AI spend. The names are deliberately not listed on this page. The connectors endpoint already returns every preset with what it pulls, the fields its setup needs and its per-status error hints, which is strictly more than a list of names, and it is current by construction.
GET /api/v1/connectors # your instances
GET /api/v1/connectors/<id> # one instance, with its materialized tables
POST /api/v1/connectors/<id>/refresh # pull now
An agent can list and refresh. It cannot create a connector, because creating one means handing over a credential, and a credential is not something a tool call should be able to conjure.
The key kind matters
The single most common setup failure has nothing to do with autario: a vendor issues several kinds of credential and only one of them can read the reporting endpoints. An inference key where the vendor wants an admin key, a project key where it wants an organization key, or a personal token where it wants a service account, all answer with an authorization error and leave the table empty.
Each connector's error text names the exact page on the vendor's site where the right key is created, and the setup form repeats it in the field help. If a connector reports an authorization failure, read that text before checking anything else.
Reading what a connector produced
Each report becomes its own table, and each table is a dataset with an id. List the connector, take the dataset ids out of its datasets array, then query them like any other dataset. There is no connector-specific query language and no second API to learn.
GET /api/v1/datasets/<dataset_id>/query?limit=50
On a plan that includes scheduled refresh (Scheduled refresh in What each plan includes), new instances refresh daily by default; on a plan without it a connection refreshes when someone presses Refresh now. A schedule of zero means manual only, which is a legitimate choice for a source you pull rarely, but it is worth knowing that it means nothing will happen unless someone presses refresh.
Connectors in an app
An app that works on a user's own connected data declares the datasets it consumes in its manifest, and the bridge enforces that list. From the app's point of view the connected table is simply a dataset id it is allowed to read. From the user's point of view, installing your app never means handing you a key.