Connect Timescale Cloud to Excel, Google Sheets, AI and REST

Set the connection mode before you save, not after

This is the one instruction on the page that matters more than it looks, and the reason is worth a paragraph.

• Timescale runs both of its pool modes on the same pooler port, and selects between them by which pool you connect to — the database name, not the address. So unlike a service that gives the pool its own port, there is nothing in the connection string for the card to read the mode from.
• That means the Connection Mode selector is a genuine choice rather than a confirmation. Pasting a pooler URL does not move it, because the URL does not carry the answer.
• The selector is what tells the agent to adapt: on the pool it turns off automatic prepared statements and skips the session reset, both of which are wrong to do through a transaction pooler.
• Set it when you create the connection. This is the point in the flow where the choice is in front of you, and it is much easier than revisiting it later.
• If you are not using the pool, leave it on direct and take the service port from the console. Nothing else changes.

TimescaleDB does the time-series work, and it does it on the server

A reasonable question about any time-series database is how much of it survives the trip into a spreadsheet. Here the answer is all of it, because none of it happens at this end.

• The connection is PostgreSQL on the wire. The agent carries the Npgsql driver and holds no TimescaleDB-specific code at all.
• So a query against a hypertable or a continuous aggregate is planned and executed by TimescaleDB itself. Chunk exclusion, compression and the aggregate machinery run on your service and what comes back is rows.
• That is the useful property rather than a limitation: time_bucket, the aggregation and the window functions all evaluate server-side, so Microsoft Excel and Google Sheets receive a finished result instead of raw readings to summarise locally.
• TLS is required and the agent enforces it for Timescale hostnames, so clearing the SSL checkbox does not produce an unencrypted connection. The certificate chain is not verified — read it as protection against anyone reading the traffic, not proof of which server answered.
• Everything the agent runs is read-only. Your credentials stay on the machine you installed it on and Query Streams keeps no copy of the data.

-- time_bucket runs on the server; Excel receives the finished series
SELECT   time_bucket('1 hour', reading_at) AS hour,
         device_id,
         AVG(temperature) AS avg_temp,
         MAX(temperature) AS peak_temp
FROM     public.sensor_readings
WHERE    reading_at >= now() - interval '7 days'
GROUP BY hour, device_id
ORDER BY hour DESC, device_id;

Timescale, Tiger Cloud, TigerData — same thing

The naming changed in 2025 and the old names are still what most people search for, so it is worth saying which is which.

• TimescaleDB is the PostgreSQL extension and still carries that name. It is what runs on the server.
• Tiger Cloud is the hosted service, previously called Timescale Cloud. TigerData is the company, previously Timescale Inc.
• The card is still labelled Timescale because that is the name on the extension and the name people look for. Nothing about the connection changed with the rebrand.
• Hostnames still end in .tsdb.cloud.timescale.com, which is what the agent recognises.