Downloads of AI Agent Frameworks Grew 25.7x in Two Years and Topped 108 Million in a Single Week in June 2026

Summary

Picture a roofing company that gets a quote request at 2 a.m. from someone who never visited the website. The request came from an AI agent that researched three local roofers, compared their service pages and filled in a form on the customer's behalf.

Agents like this are usually built with frameworks: code libraries that connect a language model to tools such as web search, page readers and form fillers. The most common frameworks are open source and installed from the Python Package Index, so their download counts are a rough gauge of how fast agent building is growing.

That gauge is climbing quickly. In the 21 sampled days of Q3 2026, ten agent framework packages were downloaded 256,300,461 times, 25.7x the same days in 2024. In the first week of June 2026 alone they passed 108 million downloads.

What we measured

We queried the public PyPI download log in Google BigQuery (bigquery-public-data.pypi.file_downloads), which records one row for every file downloaded from the Python Package Index. To keep the query affordable we sampled the first seven days of every month from January 2022 through September 2026, 399 days in total, and summed downloads by month, package, country, CI flag, installer and weekday versus weekend into one working table. A quarter in this study therefore means 21 sampled days (days 1 to 7 of each of its three months), not a full quarter. We removed downloads made by mirroring and proxy tools (bandersnatch, Nexus, Artifactory, devpi), which copy packages in bulk rather than install them. To control for the growth of PyPI itself we also pulled total daily download counts for the whole index from the table's partition metadata for the same sampled days.

We counted ten top-level agent and retrieval framework packages: langchain, langgraph, openai-agents, crewai, pydantic-ai, llama-index, autogen-agentchat, pyautogen, haystack-ai, smolagents. We left out helper packages such as langchain-community and llama-index-core, which are usually installed alongside a main framework, to avoid counting one install twice. Some frameworks still pull others in as dependencies, so group totals can overlap.

Agent framework downloads keep climbing

Each point is the first seven days of the month.

The group crossed 10 million weekly downloads in the Feb 2025 sample week and stayed above that level. The September 2026 week (53,665,671) was lower than August, in line with a 38.1% drop across all of PyPI that week.

Download share by framework

Shares are download counts only and say nothing about which framework is better.

LangChain's core package is also a dependency of many other tools, which lifts its count. Newer frameworks launched in 2024 and 2025 already account for millions of downloads: the OpenAI Agents SDK had 18,555,619 and Pydantic AI 7,495,680 in Q3 2026.

More agent builds now run through automated pipelines

CI and installer data show how these frameworks are being deployed.

Weekend downloads ran at 79.0% of the weekday pace in Q3 2026, up from 52.6% in Q3 2024, which fits more installs happening on servers that run every day. US IP addresses accounted for 87.9% of downloads, which mostly reflects where cloud servers run. For comparison, Model Context Protocol packages, which connect agents to outside tools and data, had 262,006,616 downloads in the same days.

What this means

Every agent built on these frameworks is a potential visitor to your website, and many of them act for a real customer. Agents are more likely to pick businesses whose pages state the basics plainly: what you do, where, for whom, at what price range and how to get in touch. If that information is scattered across PDFs, images or hard-to-find pages, an agent may skip you for a competitor whose site is easier to read. Moving those basics onto plain HTML pages is cheap work for an AI SEO agency and worth a lot to an agent.

Make your core service pages clear and complete, then check how AI tools see them. Our free AI visibility audit shows what AI assistants say about your business today, and our monthly LLM optimization keeps those core pages readable as agents change.

Limits

A download is not a user. One developer, one Docker build or one CI job can download the same package many times, and a single busy pipeline can move a package's numbers sharply from month to month.

Figures cover the first seven days of each month only (21 days per quarter), so they are sample totals, not full-month or full-quarter totals. Growth rates compare like-for-like sample periods.

PyPI as a whole grew from 12.1 billion downloads in the Q3 2022 sample days to 115.0 billion in the Q3 2026 sample days (9.5x), so part of every package's growth is the index growing. We report share of all PyPI downloads where it matters.

PyPI-wide downloads in the September 2026 sample week were 38.1% lower than in the August 2026 sample week (28.7 billion vs 46.4 billion), so September figures for individual packages should be read with that drop in mind.

The CI flag is only populated in this dataset from March 2024, and it only catches environments that announce themselves as CI. Automated downloads that do not set a CI variable are counted as non-CI.

Country codes come from the downloading IP address. Cloud servers sit mostly in the United States, so country splits describe where machines run, not where developers live.

Framework downloads include developers experimenting, tests and dependency installs; they are not a count of live agents.

LangChain's core package is a dependency of many other libraries, which lifts its share.

Frequently asked questions

Related studies

• Downloads of the tools that read websites for AIWeb-to-Text Tool Downloads Grew 71x Since 2022
• Downloads of the tools that read websites for AIMCP Downloads Now Exceed 10 Agent Frameworks Combined
• Downloads of the tools that read websites for AIAI Browser Agent Downloads Grew 54x in a Year
• Downloads of the tools that read websites for AIAI Web Readers Outgrew Python SEO Tools 11 to 1