Connect a database, app or cloud drive to Mammoth

Before you start

• Role: you need to be a Project Admin to create a connection. A Project Editor can use connections the project already has (see If you can't add connections).
• Plan: the launcher shows the connectors your plan includes.
• Firewall: if your database or server sits behind a firewall, allow Mammoth's IP address first. See Allow Mammoth through your firewall.

Open the Connectors launcher

1. Go to the Data Library and click New dataset.
2. In Where do you want to add data from?, choose Connectors.

You can also go to Data → Connections and click New Connection, which opens the same launcher on its own.

The launcher has a search box (Search sources or paste a link) and sources grouped in a left-hand menu:

Search the launcher by name, sentence or link

The search box reads what you type in one of three ways. The first result is what Enter does.

Known API sources

Besides its built-in connectors, the launcher lists several hundred API sources Mammoth already knows (for example payment, payroll, school or marketing systems). They appear under the matching category, or under APIs when none fits. Clicking one opens the connector workspace, where the agent reads the vendor's API documentation and builds the connector with your key. See REST API.

Connectors not on your plan

The launcher shows only the connectors your plan includes. If you search for one that isn't, it still appears, greyed out, with Not on your plan · See plans. Clicking it opens the plan picker. A connector your plan has retired stays available only while your project still has connections to it.

Connect the source

What happens when you click a source depends on its type:

• Databases and warehouses (PostgreSQL, MySQL, SQL Server, Oracle Database, Amazon Redshift, Snowflake, Google BigQuery, Databricks, Apache Hadoop, SAP HANA, MongoDB): the connection form opens first. Fill in the fields listed on the connector's page and click Connect. Once Mammoth has connected, the connector workspace opens on your tables.
• Every other source (cloud apps, drives, file servers): the connector workspace opens straight away with the Chat open. The chat starts by connecting the source and shows a card that asks for the connection details, or a button to sign in to the service. Fill it in or sign in.
• An existing connection under Your connections: the workspace opens on that connection, with nothing to re-enter.

Paste connection details into a database form

On a database connection form, click Paste connection details to fill the fields from what you copied: a connection string (such as postgresql://user@host/db), a JDBC URL, .env lines, or Host: and Port: lines from a note. SSH tunnel details fill in too.

• Mammoth shows "Filled 4 fields from your clipboard" and highlights the fields it filled. Click Undo to put back what was there.
• The text is read in your browser and is not sent anywhere. Nothing connects until you click Connect.
• If your browser blocks clipboard access, click the Host field and paste: Mammoth splits the details across the fields.
• If it finds nothing it recognises: "Your clipboard has no connection details we recognise."

If you can't add connections

Only Project Admins create connections. A Project Editor sees New sources are added by this project's admins. and can click Copy a request to copy a note for an admin, or Upload an export instead. Editors can still start a dataset from a connection the project already has. See Manage saved connections.

Work in the connector workspace

The connector workspace is a page in the Data Library (Data → Connections) with three parts. On a narrow screen they appear as the tabs Tables, Your table and Chat. On a wide screen they sit side by side.

Tables: what the connection can read

The list of tables, objects, endpoints or files the connection can read. Use Filter tables to narrow it.

• Database: click a table to open its rows with the query that read them above. See Query a database and choose how it refreshes.
• Cloud app (CRM, marketing, accounting, analytics): click a table to see its columns and sample rows, then tick the columns you want. Ticked columns gather in Your table.
• Drive or file server: browse the folders and open a file. A spreadsheet opens to its tabs. If a file can't open in the panel yet, ask the agent in the chat to import it.

Each table shows its status:

Find a missing table with Refresh list

If a table you expect is missing:

1. Click Refresh list.
2. In Refresh the list of HubSpot tables, optionally type the tables you want under Tables to look for (for example, invoices).
3. Click Refresh.

The agent looks in the source's documentation for more tables, then reads each one with your key. You can keep working meanwhile. New tables appear under Just found. If nothing new turns up: "Nothing new found. The chat says why." You can also click Ask the agent to find them or Copy a note for your admin.

Your table: the dataset you are building

Your table has one column for each column you ticked. Each column header shows which table it comes from. Click a column's name to rename it. Click Preview 100 rows to see sample rows before you save.

Join columns from a second table

When you tick columns from a second table, Mammoth joins it to the main table on a shared key. A line under Your table says which table is joined and on which key, such as Owners joined on Owner ID, and how many rows match, for example 812 of 830 rows matched.

• If only part of the table has been read, the line says the save joins every row.
• Click Change to pick another key. Each option shows how many rows it would match.
• If no column links the tables, you see, for example, "Owners isn’t joined to the main source yet".

Start from a sample table

If Your table is empty, click Use a sample table, or drop a CSV, TSV or Excel file with a header row that is shaped like the table you want. Mammoth looks for each of its columns in the source, by name and by matching values, and shows how many it found, such as 12 of 14 columns of HubSpot found. Columns it can't find are marked not found.

When a column can't be found

If a column you want can't be read, a card explains why, such as Amount: not found., or that your key can't reach that part of the source. Your options:

Columns filled from a file stay as uploaded; the other columns refresh from the source.

Chat: ask the agent for tables and columns

This connector's chat answers questions about what a source returns and can pick tables and columns for you. Ask it what you want in the table. Clicks you make in Tables show up in the chat. On a database, the chat starts closed: click Ask AI to open it.

Save the data as a dataset

1. Click Save as dataset.
2. Enter a Dataset name and pick the folder under Save in.
3. Choose how the dataset stays current under Keep it up to date (see the table below).
4. Click Save. For a database query, the button in the dialog is Save as dataset.

Mammoth reads the source in full, which can take a few minutes for a large table. You can leave the page while it saves. When it finishes, click Open dataset.

For a database query, choosing Daily at 06:00 creates a Dataset Refresh automation named after the dataset, such as Daily refresh · Orders, whose first run is the next morning. You can change its schedule later under Data → Automation.

Pick up a connector you didn't finish

A connector you open but don't save stays as a draft. The launcher lists drafts under In progress with their state:

• Not connected yet
• Pick a table
• 5 columns picked · not saved yet
• 2 columns still to find

Click a draft to reopen it. Up to two drafts show; the rest, and drafts older than 14 days, sit behind a "more drafts" line. Drafts also appear on the Connections page, where a Project Admin can Discard one. Discarding removes the draft and its chat; nothing in your data changes.

Connections that need attention, such as an expired key, appear under Needs attention with a reason (key expired or unreachable). Click one to open the connection on the Data → Connections page. Only Project Admins see this list.

Start another dataset from a saved connection

• Click the connection under Your connections in the launcher, or
• go to Data → Connections, click the connection, and click Create dataset.

A new draft opens on that connection with nothing to re-enter. If a draft can't be started, Mammoth opens the connection's own page instead, listing its Tables and endpoints; click New dataset from this there.

Connectors you can use, by category

Each connector has its own reference page with the fields to fill in and anything specific to that source.

Databases and warehouses

These connectors use direct credentials (hostname, port, username, password) and most need network access from Mammoth's servers.

Cloud storage, file servers and email

Supported file types vary by connector. A file you save from these sources (except Google Sheets and Azure Blob Storage) is saved once. To bring in new files on a schedule, use a File Collection automation.

CRM and marketing

Analytics, advertising and streaming

E-commerce, accounting and other sources

The launcher also lists several hundred known API sources by category (including APIs and Education), each built the same way as the REST API connector.

What the columns mean:

• AI query: writes SQL from a plain-language description (Write SQL in a database's SQL view).
• Schemas (or schemas/profiles): you pick a schema or profile after you sign in.
• Needs firewall rule: you must allow Mammoth's IP address, shown on the connection form, in your firewall. See Allow Mammoth through your firewall.

Troubleshooting connectors

Related

• Query a database and choose how it refreshes: the SQL view, Write SQL and refresh options.
• Manage saved connections: see, reuse, edit and delete connections, and who can do what.
• Allow Mammoth through your firewall: needed for database and SFTP connectors.
• Refresh a dataset: keep connector datasets up to date.
• Extract tables from PDFs: not a connector, but pulls tables from files already in your Data Library.
• Speed up pipelines and connectors: incremental refresh, query filtering and scheduling.