
Perplexity cannot query Google Trends on its own — it can read articles about it, but not the data. TrendFlow is a hosted MCP server you add as a custom remote connector, and it gives Perplexity the real thing: search interest over time, interest by region, related and rising queries, and what is trending right now in any country.
Perplexity moves these labels around from time to time. If the wording has changed, the shape has not: it is a custom connector, of the remote kind, pointed at an HTTPS URL. Their own walkthrough is at Adding custom remote connectors.
| Field | Value |
|---|---|
| Name | TrendFlow |
| MCP server URL | https://trendflow.mory.dev/mcp |
| Description | Google Trends: search interest over time, by region, and rising queries. |
| Transport | Streamable HTTP (under Advanced) |
| Authentication | None for the free tools, or API Key for everything |
Choose API Key as the authentication method and paste a TrendFlow key. Perplexity sends it as an Authorization: Bearer header, which is what the endpoint expects — there is nothing to configure on our side.
Create one at trendflow.mory.dev/connect. Signing in makes a free account if you do not have one, and the key looks like tfk_…. It is shown once and stored only as a hash, so copy it before you close the page.
Perplexity also supports OAuth 2.0 for connectors, and TrendFlow advertises an authorization server at /.well-known/oauth-protected-resource with dynamic client registration and PKCE. The API key is the shorter path, and it is the one we would start with.
| Ask | What it calls |
|---|---|
| What is trending in Germany right now? | get_trending_now — free |
| Compare interest in Ozempic, Wegovy, and Mounjaro over five years. | get_interest_over_time |
| Which US states search for pickleball the most? | get_interest_by_region |
| What searches are rising alongside electric bikes in the UK? | get_related_queries |
| Give me the full picture on one topic. | research_trend — everything above, in one call |
The full tool list, the filters each one takes, and what the tiers cover are in the MCP server documentation.
Every figure is normalized relative interest, scaled 0–100 within a single result set — never absolute search volume. Two separate queries are not comparable to each other. It is worth saying so in the thread, because a model asked how many people searched for something will otherwise narrate the 0–100 scale as if it were a count.
Not on its own. Perplexity can read pages about Google Trends, but it cannot query the Trends data itself. Adding TrendFlow as a custom remote MCP connector gives Perplexity live access to search interest over time, interest by region, related and rising queries, and trending searches by country.
Custom remote connectors are a paid feature: Perplexity Pro, Max, and Enterprise. They are not available on the free plan. On Enterprise, an admin must first turn on custom connectors for the organization.
No, not to start. Topic lookup, region lookup, and trending searches are free and need no account, so you can add the connector with authentication set to None and use it immediately. The historical tools — interest over time, interest by region, related queries — run on a monthly allowance and need a free TrendFlow account and an API key.
Choose API Key as the authentication method when you add the connector, and paste your TrendFlow key. Perplexity sends it as an Authorization: Bearer header, which is exactly what the TrendFlow endpoint expects. Create a key at trendflow.mory.dev/connect.
Streamable HTTP. TrendFlow serves MCP over Streamable HTTP at https://trendflow.mory.dev/mcp, over HTTPS, which is what Perplexity requires for a remote connector.
A connector has to be enabled for the thread as well as added to your account. Turn it on in the thread, then ask a question that clearly needs the data — naming Google Trends, search interest, or a comparison between terms is usually enough.
No. Every figure Google Trends returns is normalized relative interest, scaled 0 to 100 within a single result set, never an absolute count of searches. Two separate queries are not comparable to each other.