PyPI download counts already exist through pypistats and public BigQuery tables. Those are raw installs. Trends API exposes the same ecosystem as source python (not pypi) with a 0-100 score and volume when the source has it. The keyword is the exact project name. body is a JSON string and must be parsed a second time. Docs: the API reference. Caps: pricing. There is no separate /trends/pypi guide. Use the request below.
pypistats and BigQuery dumps
Those tools answer "how many pip installs." They do not answer "where does this sit on the same axis as npm and Google." Loading a BigQuery dump into a warehouse and then inventing a min-max scale in SQL creates a private score that will not match this API.
source python, not the string pypi
POST https://api.trendsapi.ai/api
Authorization: Bearer <api_key>
Content-Type: application/json
{"mode":"get_growth","source":"python","keyword":"pandas","percent_growth":["3M","12M"]}
A live pull on 2026-08-27 for pandas on python returned 3M growth of +7.04 percent (89.7 vs 83.8) with recent_volume 181419702, and 12M growth of +107.64 percent (89.7 vs 43.2) with baseline_volume 94643779. recent_date on those rows was 2026-08-17.
Project name is the normalized name
Right: pandas, requests. Wrong: Pandas, Requests, pandas-python. Copy the name from the pypi.org project page. If the inner payload is empty, the name is wrong. Do not send an import path (sklearn versus the project that provides it) unless that is the published project name.
pandas score versus install count
89.7 is the score. 181419702 is volume. A dashboard that plots 89.7 next to npm's 59.8 is comparing scores. A dashboard that plots 181 million next to npm's 102 million is comparing raw installs inside two ecosystems. Do not mix those charts. The product MCP server at https://api.trendsapi.ai/mcp expects source python for this series.