RecipeAgents / Rankings / AI Python packages

The most-used AI Python packages (October 2026)

76 AI packages on PyPI, ranked by monthly downloads and by whether that usage holds up after 3 and 6 months. Downloads aren't users: CI servers, Docker builds and other packages pulling these in all count.

1. #1

   openai

   openai

   AI and Language Models

   Widely used and holding or growing.

   Downloads / month

   341.8M

   Still used after 6 months

   Yes, 100%

   Score

   98/100

   Vendor-backed PyPI · GitHub

2. #2

   tiktoken

   tiktoken

   AI and Language Models

   Widely used and holding or growing.

   Downloads / month

   205.0M

   Still used after 6 months

   Yes, 100%

   Score

   96/100

   Vendor-backed PyPI · GitHub

3. #3

   mcp

   mcp

   AI and Language Models

   Widely used and holding or growing.

   Downloads / month

   274.2M

   Still used after 6 months

   Yes, 100%

   Score

   95/100

   Vendor-backed PyPI · GitHub

All 76 ranked packages

One row per project: packages from the same GitHub repo are folded into the most-downloaded one. Tap a row for the daily downloads chart, the churn chart and links.

1. 1openaiAI and Language ModelsVendor-backedDownloads / month341,755,176Retention kept after 3, 6 months91% / 100%StickyChurn lost after 3, 6 months9% / 0%ProblemsLowScore / 10098

   The official Python library for the openai API

   Loading chart…

   Loading chart…

   Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

   • Still there after 3 months: 91% of the download level
   • Still there after 6 months: 100%
   • Download churn (level lost) after 3 / 6 months: 9% / 0%
   • Change vs 6 months ago: +25%
   • GitHub issues opened in the last 12 months: 195
   • GitHub stars: 31,983
   • Last release: 2026-10-28, last code push 2026-10-30
   • Score parts: usage 100, staying 95, problems 94, upkeep 100

   PyPI · GitHub

2. 2tiktokenAI and Language ModelsVendor-backedDownloads / month205,025,898Retention kept after 3, 6 months91% / 100%StickyChurn lost after 3, 6 months9% / 0%ProblemsLowScore / 10096

   tiktoken is a fast BPE tokeniser for use with OpenAI's models

   Loading chart…

   Loading chart…

   Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

   • Still there after 3 months: 91% of the download level
   • Still there after 6 months: 100%
   • Download churn (level lost) after 3 / 6 months: 9% / 0%
   • Change vs 6 months ago: +40%
   • GitHub issues opened in the last 12 months: 60
   • GitHub stars: 19,557
   • Last release: 2026-08-17, last code push 2026-08-17
   • Score parts: usage 95, staying 95, problems 95, upkeep 100

   PyPI · GitHub

3. 3mcpAI and Language ModelsVendor-backedDownloads / month274,237,768Retention kept after 3, 6 months92% / 100%StickyChurn lost after 3, 6 months8% / 0%ProblemsLowScore / 10095

   Model Context Protocol SDK

   Loading chart…

   Loading chart…

   Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

   • Still there after 3 months: 92% of the download level
   • Still there after 6 months: 100%
   • Download churn (level lost) after 3 / 6 months: 8% / 0%
   • Change vs 6 months ago: +14%
   • GitHub issues opened in the last 12 months: 605
   • GitHub stars: 24,710
   • Last release: 2026-10-02, last code push 2026-10-27
   • Score parts: usage 98, staying 96, problems 78, upkeep 100

   PyPI · GitHub

4. 4anthropicAI and Language ModelsVendor-backedDownloads / month186,460,190Retention kept after 3, 6 months100% / 100%StickyChurn lost after 3, 6 months0% / 0%ProblemsLowScore / 10095

   The official Python library for the anthropic API

   Loading chart…

   Loading chart…

   Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

   • Still there after 3 months: 100% of the download level
   • Still there after 6 months: 100%
   • Download churn (level lost) after 3 / 6 months: 0% / 0%
   • Change vs 6 months ago: +71%
   • GitHub issues opened in the last 12 months: 254
   • GitHub stars: 4,022
   • Last release: 2026-10-23, last code push 2026-10-23
   • Score parts: usage 94, staying 100, problems 83, upkeep 100

   PyPI · GitHub

5. 5langchain-coreAI and Language Models · +6 related packagesVendor-backedDownloads / month152,040,588Retention kept after 3, 6 months90% / 100%HoldingChurn lost after 3, 6 months10% / 0%ProblemsLowScore / 10090

   Building applications with LLMs through composability

   Loading chart…

   Loading chart…

   Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

   • Still there after 3 months: 90% of the download level
   • Still there after 6 months: 100%
   • Download churn (level lost) after 3 / 6 months: 10% / 0%
   • Change vs 6 months ago: +40%
   • GitHub issues opened in the last 12 months: 1,847 (shared by the whole repo)
   • GitHub stars: 148,705
   • Last release: 2026-10-29, last code push 2026-10-30
   • Same project, not listed separately: langchain, langchain-anthropic, langchain-classic, langchain-groq, langchain-ollama, langchain-openai
   • Score parts: usage 92, staying 95, problems 68, upkeep 100

   PyPI · GitHub

6. 6langfuseDevOpsVendor-backedDownloads / month31,483,984Retention kept after 3, 6 months100% / 100%StickyChurn lost after 3, 6 months0% / 0%ProblemsLowScore / 10089

   Langfuse Python SDK - LLM observability/tracing, datasets, experiments, LLM-as-a-judge evaluation, and prompt management

   Loading chart…

   Loading chart…

   Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

   • Still there after 3 months: 100% of the download level
   • Still there after 6 months: 100%
   • Download churn (level lost) after 3 / 6 months: 0% / 0%
   • Change vs 6 months ago: +43%
   • GitHub issues opened in the last 12 months: 37
   • GitHub stars: 506
   • Last release: 2026-10-21, last code push 2026-10-30
   • Score parts: usage 77, staying 100, problems 85, upkeep 100

   PyPI · GitHub

7. 7langgraphAI and Language Models · +3 related packagesVendor-backedDownloads / month58,157,120Retention kept after 3, 6 months80% / 100%HoldingChurn lost after 3, 6 months20% / 0%ProblemsLowScore / 10086

   Building stateful, multi-actor applications with LLMs

   Loading chart…

   Loading chart…

   Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

   • Still there after 3 months: 80% of the download level
   • Still there after 6 months: 100%
   • Download churn (level lost) after 3 / 6 months: 20% / 0%
   • Change vs 6 months ago: +13%
   • GitHub issues opened in the last 12 months: 701 (shared by the whole repo)
   • GitHub stars: 44,145
   • Last release: 2026-10-26, last code push 2026-10-29
   • Same project, not listed separately: langgraph-checkpoint-postgres, langgraph-checkpoint-sqlite, langgraph-cli
   • Score parts: usage 83, staying 90, problems 69, upkeep 100

   PyPI · GitHub

8. 8langchain-google-vertexaiAI and Language Models · +2 related packagesVendor-backedDownloads / month30,941,092Retention kept after 3, 6 months93% / 100%StickyChurn lost after 3, 6 months7% / 0%ProblemsLowScore / 10085

   An integration package connecting Google VertexAI and LangChain

   Loading chart…

   Loading chart…

   Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

   • Still there after 3 months: 93% of the download level
   • Still there after 6 months: 100%
   • Download churn (level lost) after 3 / 6 months: 7% / 0%
   • Change vs 6 months ago: +34%
   • GitHub issues opened in the last 12 months: 170 (shared by the whole repo)
   • GitHub stars: 406
   • Last release: 2026-10-23, last code push 2026-10-29
   • Same project, not listed separately: langchain-google-community, langchain-google-genai
   • Score parts: usage 77, staying 96, problems 71, upkeep 100

   PyPI · GitHub

9. 9google-genaiAI and Language ModelsVendor-backedDownloads / month157,493,952Retention kept after 3, 6 months56% / 75%FadingChurn lost after 3, 6 months44% / 25%ProblemsLowScore / 10084

   GenAI Python SDK

   Loading chart…

   Loading chart…

   Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

   • Still there after 3 months: 56% of the download level
   • Still there after 6 months: 75%
   • Download churn (level lost) after 3 / 6 months: 44% / 25%
   • Change vs 6 months ago: −25%
   • GitHub issues opened in the last 12 months: 425
   • GitHub stars: 4,049
   • Last release: 2026-10-29, last code push 2026-10-29
   • Score parts: usage 93, staying 65, problems 82, upkeep 100

   PyPI · GitHub

10. 10agent-client-protocolAI and Language ModelsDownloads / month23,986,200Retention kept after 3, 6 months76% / 100%HoldingChurn lost after 3, 6 months24% / 0%ProblemsLowScore / 10084

    A Python implement of Agent Client Protocol (ACP, by Zed Industries)

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 76% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 24% / 0%
    • Change vs 6 months ago: +254%
    • GitHub issues opened in the last 12 months: 24
    • GitHub stars: 342
    • Last release: 2026-10-12, last code push 2026-10-29
    • Score parts: usage 75, staying 88, problems 86, upkeep 100

    PyPI · GitHub

11. 11openai-whisperMediaVendor-backedDownloads / month5,900,328Retention kept after 3, 6 months100% / 100%StickyChurn lost after 3, 6 months0% / 0%ProblemsLowScore / 10084

    Robust Speech Recognition via Large-Scale Weak Supervision

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 100% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 0% / 0%
    • Change vs 6 months ago: +172%
    • GitHub issues opened in the last 12 months: 0
    • GitHub stars: 111,569
    • Last release: 2025-06-26, last code push 2026-08-31
    • Score parts: usage 61, staying 100, problems 100, upkeep 100

    PyPI · GitHub

12. 12claude-agent-sdkAI and Language ModelsVendor-backedDownloads / month34,425,524Retention kept after 3, 6 months100% / 100%StickyChurn lost after 3, 6 months0% / 0%ProblemsMediumScore / 10083

    Python SDK for Claude Code

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 100% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 0% / 0%
    • Change vs 6 months ago: +202%
    • GitHub issues opened in the last 12 months: 456
    • GitHub stars: 8,392
    • Last release: 2026-10-30, last code push 2026-10-30
    • Score parts: usage 78, staying 100, problems 46, upkeep 100

    PyPI · GitHub

13. 13groqAI and Language ModelsVendor-backedDownloads / month16,923,984Retention kept after 3, 6 months64% / 100%HoldingChurn lost after 3, 6 months36% / 0%ProblemsLowScore / 10083

    The official Python library for the groq API

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 64% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 36% / 0%
    • Change vs 6 months ago: −0%
    • GitHub issues opened in the last 12 months: 0
    • GitHub stars: 622
    • Last release: 2026-08-26, last code push 2026-09-04
    • Score parts: usage 71, staying 82, problems 100, upkeep 100

    PyPI · GitHub

14. 14fastmcpAI and Language ModelsVendor-backedDownloads / month61,412,988Retention kept after 3, 6 months77% / 90%HoldingChurn lost after 3, 6 months23% / 10%ProblemsMediumScore / 10082

    The fast, Pythonic way to build MCP servers and clients.

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 77% of the download level
    • Still there after 6 months: 90%
    • Download churn (level lost) after 3 / 6 months: 23% / 10%
    • Change vs 6 months ago: −10%
    • GitHub issues opened in the last 12 months: 1,267
    • GitHub stars: 28,242
    • Last release: 2026-10-27, last code push 2026-10-30
    • Score parts: usage 84, staying 83, problems 60, upkeep 100

    PyPI · GitHub

15. 15mini-swe-agentAI and Language ModelsDownloads / month15,614,760Retention kept after 3, 6 months100% / 100%StickyChurn lost after 3, 6 months0% / 0%ProblemsMediumScore / 10082

    Mini SWE Agent - A simple AI software engineering agent

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 100% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 0% / 0%
    • Change vs 6 months ago: +805%
    • GitHub issues opened in the last 12 months: 129
    • GitHub stars: 8,967
    • Last release: 2026-07-23, last code push 2026-10-26
    • Score parts: usage 70, staying 100, problems 58, upkeep 100

    PyPI · GitHub

16. 16transformersAI and Language ModelsVendor-backedDownloads / month106,582,658Retention kept after 3, 6 months62% / 82%HoldingChurn lost after 3, 6 months38% / 18%ProblemsMediumScore / 10081

    Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 62% of the download level
    • Still there after 6 months: 82%
    • Download churn (level lost) after 3 / 6 months: 38% / 18%
    • Change vs 6 months ago: −18%
    • GitHub issues opened in the last 12 months: 1,628
    • GitHub stars: 168,343
    • Last release: 2026-10-30, last code push 2026-10-30
    • Score parts: usage 89, staying 72, problems 57, upkeep 100

    PyPI · GitHub

17. 17strands-agentsAI and Language ModelsVendor-backedDownloads / month36,418,032Retention kept after 3, 6 months100% / 100%StickyChurn lost after 3, 6 months0% / 0%ProblemsHighScore / 10081

    A model-driven approach to building AI agents in just a few lines of code

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 100% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 0% / 0%
    • Change vs 6 months ago: +570%
    • GitHub issues opened in the last 12 months: 1,367
    • GitHub stars: 9,001
    • Last release: 2026-10-21, last code push 2026-10-30
    • Score parts: usage 79, staying 100, problems 28, upkeep 100

    PyPI · GitHub

18. 18sentence-transformersData and StorageVendor-backedDownloads / month22,293,166Retention kept after 3, 6 months75% / 91%HoldingChurn lost after 3, 6 months25% / 9%ProblemsLowScore / 10079

    Embeddings, Retrieval, and Reranking

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 75% of the download level
    • Still there after 6 months: 91%
    • Download churn (level lost) after 3 / 6 months: 25% / 9%
    • Change vs 6 months ago: −9%
    • GitHub issues opened in the last 12 months: 131
    • GitHub stars: 19,236
    • Last release: 2026-10-21, last code push 2026-10-30
    • Score parts: usage 74, staying 83, problems 66, upkeep 100

    PyPI · GitHub

19. 19ollamaAI and Language ModelsVendor-backedDownloads / month12,340,678Retention kept after 3, 6 months68% / 100%HoldingChurn lost after 3, 6 months32% / 0%ProblemsLowScore / 10079

    The official Python client for Ollama.

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 68% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 32% / 0%
    • Change vs 6 months ago: +6%
    • GitHub issues opened in the last 12 months: 42
    • GitHub stars: 10,644
    • Last release: 2026-09-29, last code push 2026-10-16
    • Score parts: usage 68, staying 84, problems 75, upkeep 100

    PyPI · GitHub

20. 20e2bDevOps · +1 related packagesVendor-backedDownloads / month8,090,320Retention kept after 3, 6 months100% / 100%StickyChurn lost after 3, 6 months0% / 0%ProblemsMediumScore / 10077

    E2B SDK that give agents cloud environments

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 100% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 0% / 0%
    • Change vs 6 months ago: +115%
    • GitHub issues opened in the last 12 months: 157 (shared by the whole repo)
    • GitHub stars: 14,615
    • Last release: 2026-10-28, last code push 2026-10-29
    • Same project, not listed separately: e2b-code-interpreter
    • Score parts: usage 64, staying 100, problems 40, upkeep 100

    PyPI · GitHub

21. 21browser-use-sdkWeb and APIsVendor-backedDownloads / month5,899,628Retention kept after 3, 6 months59% / 100%FadingChurn lost after 3, 6 months41% / 0%ProblemsLowScore / 10077

    Python SDK for the Browser Use cloud API

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 59% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 41% / 0%
    • Change vs 6 months ago: +111%
    • GitHub issues opened in the last 12 months: 7
    • GitHub stars: 28
    • Last release: 2026-10-17, last code push 2026-10-30
    • Score parts: usage 61, staying 80, problems 89, upkeep 100

    PyPI · GitHub

22. 22pymupdf4llmData and StorageDownloads / month14,729,932Retention kept after 3, 6 months90% / 72%HoldingChurn lost after 3, 6 months10% / 28%ProblemsNo repo foundScore / 10075

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 90% of the download level
    • Still there after 6 months: 72%
    • Download churn (level lost) after 3 / 6 months: 10% / 28%
    • Change vs 6 months ago: −28%
    • GitHub issues opened in the last 12 months: no repo found
    • GitHub stars: none found
    • Last release: 2026-10-19
    • Score parts: usage 70, staying 81, problems 50, upkeep 100

    PyPI

23. 23qdrant-clientData and StorageVendor-backedDownloads / month14,213,626Retention kept after 3, 6 months73% / 67%HoldingChurn lost after 3, 6 months27% / 33%ProblemsMediumScore / 10074

    Client library for the Qdrant vector search engine

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 73% of the download level
    • Still there after 6 months: 67%
    • Download churn (level lost) after 3 / 6 months: 27% / 33%
    • Change vs 6 months ago: −33%
    • GitHub issues opened in the last 12 months: 110
    • GitHub stars: 1,373
    • Last release: 2026-10-22, last code push 2026-10-29
    • Score parts: usage 70, staying 70, problems 65, upkeep 100

    PyPI · GitHub

24. 24lancedbData and StorageVendor-backedDownloads / month7,475,860Retention kept after 3, 6 months97% / 100%StickyChurn lost after 3, 6 months3% / 0%ProblemsHighScore / 10074

    lancedb

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 97% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 3% / 0%
    • Change vs 6 months ago: +24%
    • GitHub issues opened in the last 12 months: 481
    • GitHub stars: 11,813
    • Last release: 2026-10-29, last code push 2026-10-30
    • Score parts: usage 63, staying 99, problems 25, upkeep 100

    PyPI · GitHub

25. 25bedrock-agentcoreDevOpsVendor-backedDownloads / month6,952,512Retention kept after 3, 6 months84% / 100%HoldingChurn lost after 3, 6 months16% / 0%ProblemsMediumScore / 10073

    An SDK for using Bedrock AgentCore

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 84% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 16% / 0%
    • Change vs 6 months ago: +12%
    • GitHub issues opened in the last 12 months: 142
    • GitHub stars: 781
    • Last release: 2026-10-27, last code push 2026-10-29
    • Score parts: usage 63, staying 92, problems 37, upkeep 100

    PyPI · GitHub

26. 26openinference-instrumentation-langchainDevOps · +2 related packagesVendor-backedDownloads / month3,592,974Retention kept after 3, 6 months99% / 100%StickyChurn lost after 3, 6 months1% / 0%ProblemsMediumScore / 10073

    OpenInference LangChain Instrumentation

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 99% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 1% / 0%
    • Change vs 6 months ago: +54%
    • GitHub issues opened in the last 12 months: 429 (shared by the whole repo)
    • GitHub stars: 1,295
    • Last release: 2026-10-29, last code push 2026-10-30
    • Same project, not listed separately: openinference-instrumentation-openai, openinference-instrumentation-openai-agents
    • Score parts: usage 56, staying 99, problems 38, upkeep 100

    PyPI · GitHub

27. 27langchain-postgresData and StorageDownloads / month1,905,120Retention kept after 3, 6 months100% / 100%StickyChurn lost after 3, 6 months0% / 0%ProblemsNo repo foundScore / 10073

    An integration package connecting Postgres and LangChain

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 100% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 0% / 0%
    • Change vs 6 months ago: +50%
    • GitHub issues opened in the last 12 months: no repo found
    • GitHub stars: none found
    • Last release: 2026-10-06
    • Score parts: usage 50, staying 100, problems 50, upkeep 100

    PyPI

28. 28deepagentsAI and Language ModelsVendor-backedDownloads / month8,022,490Retention kept after 3, 6 months100% / 100%StickyChurn lost after 3, 6 months0% / 0%ProblemsHighScore / 10072

    Production-ready, extensible agent harness with a built-in filesystem and context management, sub-agent delegation, skills, and long-term memory.

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 100% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 0% / 0%
    • Change vs 6 months ago: +147%
    • GitHub issues opened in the last 12 months: 1,296
    • GitHub stars: 30,780
    • Last release: 2026-10-23, last code push 2026-10-30
    • Score parts: usage 64, staying 100, problems 8, upkeep 100

    PyPI · GitHub

29. 29voyageaiData and StorageDownloads / month5,101,600Retention kept after 3, 6 months100% / 100%StickyChurn lost after 3, 6 months0% / 0%ProblemsNo repo foundScore / 10072

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 100% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 0% / 0%
    • Change vs 6 months ago: +104%
    • GitHub issues opened in the last 12 months: no repo found
    • GitHub stars: none found
    • Last release: 2026-07-10
    • Score parts: usage 60, staying 100, problems 50, upkeep 70

    PyPI

30. 30langchain-awsAI and Language ModelsVendor-backedDownloads / month10,159,436Retention kept after 3, 6 months72% / 77%HoldingChurn lost after 3, 6 months28% / 23%ProblemsMediumScore / 10071

    An integration package connecting AWS and LangChain

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 72% of the download level
    • Still there after 6 months: 77%
    • Download churn (level lost) after 3 / 6 months: 28% / 23%
    • Change vs 6 months ago: −23%
    • GitHub issues opened in the last 12 months: 155
    • GitHub stars: 353
    • Last release: 2026-10-26, last code push 2026-10-29
    • Score parts: usage 66, staying 75, problems 49, upkeep 100

    PyPI · GitHub

31. 31llama-index-coreAI and Language Models · +13 related packagesVendor-backedDownloads / month6,529,250Retention kept after 3, 6 months43% / 92%FadingChurn lost after 3, 6 months57% / 8%ProblemsMediumScore / 10070

    Interface between LLMs and your data

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 43% of the download level
    • Still there after 6 months: 92%
    • Download churn (level lost) after 3 / 6 months: 57% / 8%
    • Change vs 6 months ago: −8%
    • GitHub issues opened in the last 12 months: 707 (shared by the whole repo)
    • GitHub stars: 52,741
    • Last release: 2026-09-21, last code push 2026-10-26
    • Same project, not listed separately: llama-index, llama-index-agent-openai, llama-index-cli, llama-index-embeddings-bedrock, llama-index-embeddings-openai, llama-index-indices-managed-llama-cloud, llama-index-legacy, llama-index-llms-openai and 5 more
    • Score parts: usage 62, staying 67, problems 63, upkeep 100

    PyPI · GitHub

32. 32dspyAI and Language ModelsVendor-backedDownloads / month5,782,448Retention kept after 3, 6 months89% / 91%HoldingChurn lost after 3, 6 months11% / 9%ProblemsHighScore / 10070

    DSPy

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 89% of the download level
    • Still there after 6 months: 91%
    • Download churn (level lost) after 3 / 6 months: 11% / 9%
    • Change vs 6 months ago: −9%
    • GitHub issues opened in the last 12 months: 366
    • GitHub stars: 38,852
    • Last release: 2026-10-12, last code push 2026-10-29
    • Score parts: usage 61, staying 90, problems 23, upkeep 100

    PyPI · GitHub

33. 33livekit-agentsCommunication · +1 related packagesVendor-backedDownloads / month4,590,152Retention kept after 3, 6 months100% / 100%StickyChurn lost after 3, 6 months0% / 0%ProblemsHighScore / 10070

    A powerful framework for building realtime voice AI agents

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 100% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 0% / 0%
    • Change vs 6 months ago: +31%
    • GitHub issues opened in the last 12 months: 993 (shared by the whole repo)
    • GitHub stars: 15,003
    • Last release: 2026-10-28, last code push 2026-10-30
    • Same project, not listed separately: livekit-plugins-openai
    • Score parts: usage 59, staying 100, problems 11, upkeep 100

    PyPI · GitHub

34. 34mcp-proxy-for-awsWeb and APIsVendor-backedDownloads / month1,496,362Retention kept after 3, 6 months98% / 100%StickyChurn lost after 3, 6 months2% / 0%ProblemsMediumScore / 10069

    MCP Proxy for AWS

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 98% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 2% / 0%
    • Change vs 6 months ago: +185%
    • GitHub issues opened in the last 12 months: 38
    • GitHub stars: 366
    • Last release: 2026-09-15, last code push 2026-10-28
    • Score parts: usage 48, staying 99, problems 35, upkeep 100

    PyPI · GitHub

35. 35google-adkAI and Language ModelsVendor-backedDownloads / month10,242,344Retention kept after 3, 6 months61% / 100%HoldingChurn lost after 3, 6 months39% / 0%ProblemsHighScore / 10068

    Agent Development Kit

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 61% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 39% / 0%
    • Change vs 6 months ago: +44%
    • GitHub issues opened in the last 12 months: 1,815
    • GitHub stars: 22,047
    • Last release: 2026-10-23, last code push 2026-10-30
    • Score parts: usage 66, staying 81, problems 17, upkeep 100

    PyPI · GitHub

36. 36pineconeData and StorageVendor-backedDownloads / month4,323,102Retention kept after 3, 6 months57% / 67%FadingChurn lost after 3, 6 months43% / 33%ProblemsLowScore / 10068

    Pinecone Python SDK

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 57% of the download level
    • Still there after 6 months: 67%
    • Download churn (level lost) after 3 / 6 months: 43% / 33%
    • Change vs 6 months ago: −33%
    • GitHub issues opened in the last 12 months: 17
    • GitHub stars: 456
    • Last release: 2026-10-23, last code push 2026-10-23
    • Score parts: usage 58, staying 62, problems 74, upkeep 100

    PyPI · GitHub

37. 37mistralaiAI and Language ModelsVendor-backedDownloads / month14,804,846Retention kept after 3, 6 months30% / 48%FadingChurn lost after 3, 6 months70% / 52%ProblemsLowScore / 10067

    Python Client SDK for the Mistral AI API.

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 30% of the download level
    • Still there after 6 months: 48%
    • Download churn (level lost) after 3 / 6 months: 70% / 52%
    • Change vs 6 months ago: −52%
    • GitHub issues opened in the last 12 months: 66
    • GitHub stars: 773
    • Last release: 2026-10-28, last code push 2026-10-28
    • Score parts: usage 70, staying 39, problems 80, upkeep 100

    PyPI · GitHub

38. 38instructorAI and Language ModelsVendor-backedDownloads / month9,348,010Retention kept after 3, 6 months49% / 64%FadingChurn lost after 3, 6 months51% / 36%ProblemsMediumScore / 10067

    structured outputs for llm

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 49% of the download level
    • Still there after 6 months: 64%
    • Download churn (level lost) after 3 / 6 months: 51% / 36%
    • Change vs 6 months ago: −36%
    • GitHub issues opened in the last 12 months: 193
    • GitHub stars: 14,081
    • Last release: 2026-09-09, last code push 2026-10-26
    • Score parts: usage 66, staying 57, problems 55, upkeep 100

    PyPI · GitHub

39. 39mem0aiData and StorageDownloads / month2,457,448Retention kept after 3, 6 months68% / 86%HoldingChurn lost after 3, 6 months32% / 14%ProblemsNo repo foundScore / 10067

    Long-term memory for AI Agents

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 68% of the download level
    • Still there after 6 months: 86%
    • Download churn (level lost) after 3 / 6 months: 32% / 14%
    • Change vs 6 months ago: −14%
    • GitHub issues opened in the last 12 months: no repo found
    • GitHub stars: none found
    • Last release: 2026-09-25
    • Score parts: usage 53, staying 77, problems 50, upkeep 100

    PyPI

40. 40chromadbData and StorageVendor-backedDownloads / month8,335,600Retention kept after 3, 6 months65% / 62%HoldingChurn lost after 3, 6 months35% / 38%ProblemsMediumScore / 10066

    Chroma.

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 65% of the download level
    • Still there after 6 months: 62%
    • Download churn (level lost) after 3 / 6 months: 35% / 38%
    • Change vs 6 months ago: −38%
    • GitHub issues opened in the last 12 months: 227
    • GitHub stars: 29,609
    • Last release: 2026-05-05, last code push 2026-10-28
    • Score parts: usage 64, staying 64, problems 43, upkeep 100

    PyPI · GitHub

41. 41agnoAI and Language ModelsVendor-backedDownloads / month2,187,598Retention kept after 3, 6 months100% / 100%StickyChurn lost after 3, 6 months0% / 0%ProblemsHighScore / 10066

    The programming language for agentic software.

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 100% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 0% / 0%
    • Change vs 6 months ago: +17%
    • GitHub issues opened in the last 12 months: 1,732
    • GitHub stars: 42,965
    • Last release: 2026-10-29, last code push 2026-10-30
    • Score parts: usage 52, staying 100, problems 5, upkeep 100

    PyPI · GitHub

42. 42arize-phoenix-otelDevOpsVendor-backedDownloads / month2,185,932Retention kept after 3, 6 months100% / 100%StickyChurn lost after 3, 6 months0% / 0%ProblemsHighScore / 10066

    LLM Observability

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 100% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 0% / 0%
    • Change vs 6 months ago: +16%
    • GitHub issues opened in the last 12 months: 2,458
    • GitHub stars: 12,073
    • Last release: 2026-09-28, last code push 2026-10-30
    • Score parts: usage 52, staying 100, problems 3, upkeep 100

    PyPI · GitHub

43. 43deepevalDevOpsVendor-backedDownloads / month3,620,498Retention kept after 3, 6 months60% / 97%HoldingChurn lost after 3, 6 months40% / 3%ProblemsHighScore / 10065

    The LLM Evaluation Framework

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 60% of the download level
    • Still there after 6 months: 97%
    • Download churn (level lost) after 3 / 6 months: 40% / 3%
    • Change vs 6 months ago: −3%
    • GitHub issues opened in the last 12 months: 257
    • GitHub stars: 19,193
    • Last release: 2026-10-27, last code push 2026-10-29
    • Score parts: usage 56, staying 78, problems 26, upkeep 100

    PyPI · GitHub

44. 44braintrust-langchainDevOpsDownloads / month1,945,188Retention kept after 3, 6 months82% / 100%HoldingChurn lost after 3, 6 months18% / 0%ProblemsHighScore / 10065

    DEPRECATED: LangChain integration is now included in the main braintrust package. Install braintrust instead.

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 82% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 18% / 0%
    • Change vs 6 months ago: +19%
    • GitHub issues opened in the last 12 months: 246
    • GitHub stars: 22
    • Last release: 2026-04-01, last code push 2026-10-29
    • Score parts: usage 50, staying 91, problems 14, upkeep 100

    PyPI · GitHub

45. 45replicateAI and Language ModelsVendor-backedDownloads / month1,683,934Retention kept after 3, 6 months99% / 100%StickyChurn lost after 3, 6 months1% / 0%ProblemsMediumScore / 10065

    Python client for Replicate

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 99% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 1% / 0%
    • Change vs 6 months ago: +12%
    • GitHub issues opened in the last 12 months: 12
    • GitHub stars: 918
    • Last release: 2025-12-18, last code push 2025-08-26
    • Score parts: usage 49, staying 99, problems 62, upkeep 40

    PyPI · GitHub

46. 46k8s-agent-sandboxDevOpsDownloads / month1,371,384Retention kept after 3, 6 months100% / 100%StickyChurn lost after 3, 6 months0% / 0%ProblemsHighScore / 10065

    A client library to interact with the Agentic Sandbox on Kubernetes.

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 100% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 0% / 0%
    • Change vs 6 months ago: +820%
    • GitHub issues opened in the last 12 months: 465
    • GitHub stars: 4,450
    • Last release: 2026-10-29, last code push 2026-10-29
    • Score parts: usage 47, staying 100, problems 6, upkeep 100

    PyPI · GitHub

47. 47litellmAI and Language ModelsDownloads / month93,097,732Retention kept after 3, 6 months15% / 28%FadingChurn lost after 3, 6 months85% / 72%ProblemsMediumScore / 10064

    Library to easily interface with LLM API providers

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 15% of the download level
    • Still there after 6 months: 28%
    • Download churn (level lost) after 3 / 6 months: 85% / 72%
    • Change vs 6 months ago: −72%
    • GitHub issues opened in the last 12 months: 6,393
    • GitHub stars: 62,359
    • Last release: 2026-10-30, last code push 2026-10-30
    • Score parts: usage 88, staying 22, problems 48, upkeep 100

    PyPI · GitHub

48. 48cohereAI and Language ModelsVendor-backedDownloads / month11,549,664Retention kept after 3, 6 months30% / 31%FadingChurn lost after 3, 6 months70% / 69%ProblemsLowScore / 10064

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 30% of the download level
    • Still there after 6 months: 31%
    • Download churn (level lost) after 3 / 6 months: 70% / 69%
    • Change vs 6 months ago: −69%
    • GitHub issues opened in the last 12 months: 27
    • GitHub stars: 406
    • Last release: 2026-10-21, last code push 2026-10-27
    • Score parts: usage 68, staying 30, problems 88, upkeep 100

    PyPI · GitHub

49. 49unstructured-clientData and StorageVendor-backedDownloads / month3,817,114Retention kept after 3, 6 months42% / 38%FadingChurn lost after 3, 6 months58% / 62%ProblemsLowScore / 10064

    Python Client SDK for Unstructured API

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 42% of the download level
    • Still there after 6 months: 38%
    • Download churn (level lost) after 3 / 6 months: 58% / 62%
    • Change vs 6 months ago: −62%
    • GitHub issues opened in the last 12 months: 2
    • GitHub stars: 120
    • Last release: 2026-10-19, last code push 2026-10-21
    • Score parts: usage 57, staying 40, problems 97, upkeep 100

    PyPI · GitHub

50. 50pydantic-ai-skillsAI and Language ModelsDownloads / month1,579,704Retention kept after 3, 6 months45% / 100%FadingChurn lost after 3, 6 months55% / 0%ProblemsMediumScore / 10064

    Remote skill registries and bundled-file execution for Pydantic AI Agent Skills

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 45% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 55% / 0%
    • Change vs 6 months ago: +435%
    • GitHub issues opened in the last 12 months: 23
    • GitHub stars: 381
    • Last release: 2026-10-28, last code push 2026-10-30
    • Score parts: usage 48, staying 72, problems 52, upkeep 100

    PyPI · GitHub

51. 51openai-agentsAI and Language ModelsVendor-backedDownloads / month14,810,390Retention kept after 3, 6 months42% / 58%FadingChurn lost after 3, 6 months58% / 42%ProblemsMediumScore / 10063

    OpenAI Agents SDK

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 42% of the download level
    • Still there after 6 months: 58%
    • Download churn (level lost) after 3 / 6 months: 58% / 42%
    • Change vs 6 months ago: −42%
    • GitHub issues opened in the last 12 months: 707
    • GitHub stars: 30,389
    • Last release: 2026-10-15, last code push 2026-10-27
    • Score parts: usage 70, staying 50, problems 34, upkeep 100

    PyPI · GitHub

52. 52togetherAI and Language ModelsVendor-backedDownloads / month1,958,600Retention kept after 3, 6 months81% / 24%FadingChurn lost after 3, 6 months19% / 76%ProblemsLowScore / 10063

    The official Python library for the together API

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 81% of the download level
    • Still there after 6 months: 24%
    • Download churn (level lost) after 3 / 6 months: 19% / 76%
    • Change vs 6 months ago: −76%
    • GitHub issues opened in the last 12 months: 7
    • GitHub stars: 11
    • Last release: 2026-10-22, last code push 2026-10-28
    • Score parts: usage 51, staying 53, problems 77, upkeep 100

    PyPI · GitHub

53. 53strands-agents-toolsAI and Language ModelsVendor-backedDownloads / month2,604,014Retention kept after 3, 6 months45% / 80%FadingChurn lost after 3, 6 months55% / 20%ProblemsMediumScore / 10061

    A collection of specialized tools for Strands Agents

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 45% of the download level
    • Still there after 6 months: 80%
    • Download churn (level lost) after 3 / 6 months: 55% / 20%
    • Change vs 6 months ago: −20%
    • GitHub issues opened in the last 12 months: 77
    • GitHub stars: 1,340
    • Last release: 2026-10-21, last code push 2026-10-29
    • Score parts: usage 53, staying 63, problems 42, upkeep 100

    PyPI · GitHub

54. 54traceloop-sdkDevOpsVendor-backedDownloads / month2,038,722Retention kept after 3, 6 months76% / 76%HoldingChurn lost after 3, 6 months24% / 24%ProblemsHighScore / 10061

    Traceloop Software Development Kit (SDK) for Python

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 76% of the download level
    • Still there after 6 months: 76%
    • Download churn (level lost) after 3 / 6 months: 24% / 24%
    • Change vs 6 months ago: −24%
    • GitHub issues opened in the last 12 months: 160
    • GitHub stars: 7,512
    • Last release: 2026-09-29, last code push 2026-10-19
    • Score parts: usage 51, staying 76, problems 22, upkeep 100

    PyPI · GitHub

55. 55databricks-agentsAI and Language ModelsDownloads / month3,568,684Retention kept after 3, 6 months47% / 54%FadingChurn lost after 3, 6 months53% / 46%ProblemsNo repo foundScore / 10060

    Mosaic AI Agent Framework SDK

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 47% of the download level
    • Still there after 6 months: 54%
    • Download churn (level lost) after 3 / 6 months: 53% / 46%
    • Change vs 6 months ago: −46%
    • GitHub issues opened in the last 12 months: no repo found
    • GitHub stars: none found
    • Last release: 2026-10-21
    • Score parts: usage 56, staying 50, problems 50, upkeep 100

    PyPI

56. 56unstructuredData and StorageVendor-backedDownloads / month2,842,280Retention kept after 3, 6 months51% / 50%FadingChurn lost after 3, 6 months49% / 50%ProblemsMediumScore / 10060

    A library that prepares raw documents for downstream ML tasks.

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 51% of the download level
    • Still there after 6 months: 50%
    • Download churn (level lost) after 3 / 6 months: 49% / 50%
    • Change vs 6 months ago: −50%
    • GitHub issues opened in the last 12 months: 58
    • GitHub stars: 15,644
    • Last release: 2026-10-30, last code push 2026-10-30
    • Score parts: usage 54, staying 51, problems 54, upkeep 100

    PyPI · GitHub

57. 57outlinesAI and Language ModelsVendor-backedDownloads / month1,279,796Retention kept after 3, 6 months62% / 83%HoldingChurn lost after 3, 6 months38% / 17%ProblemsMediumScore / 10060

    Probabilistic Generative Model Programming

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 62% of the download level
    • Still there after 6 months: 83%
    • Download churn (level lost) after 3 / 6 months: 38% / 17%
    • Change vs 6 months ago: −17%
    • GitHub issues opened in the last 12 months: 68
    • GitHub stars: 15,988
    • Last release: 2026-08-06, last code push 2026-09-21
    • Score parts: usage 46, staying 72, problems 32, upkeep 100

    PyPI · GitHub

58. 58browser-useWeb and APIsVendor-backedDownloads / month6,105,848Retention kept after 3, 6 months14% / 65%FadingChurn lost after 3, 6 months86% / 35%ProblemsMediumScore / 10059

    Open-source browser agent + hosted cloud browsers for AI agents, from $0.02/browser-hour

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 14% of the download level
    • Still there after 6 months: 65%
    • Download churn (level lost) after 3 / 6 months: 86% / 35%
    • Change vs 6 months ago: −35%
    • GitHub issues opened in the last 12 months: 507
    • GitHub stars: 119,542
    • Last release: 2026-10-26, last code push 2026-10-30
    • Score parts: usage 61, staying 39, problems 51, upkeep 100

    PyPI · GitHub

59. 59azure-ai-agentsAI and Language ModelsDownloads / month4,678,282Retention kept after 3, 6 months61% / 64%HoldingChurn lost after 3, 6 months39% / 36%ProblemsHighScore / 10059

    Microsoft Corporation Azure AI Agents Client Library for Python

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 61% of the download level
    • Still there after 6 months: 64%
    • Download churn (level lost) after 3 / 6 months: 39% / 36%
    • Change vs 6 months ago: −36%
    • GitHub issues opened in the last 12 months: 1,070
    • GitHub stars: 5,624
    • Last release: 2025-10-24, last code push 2026-10-30
    • Score parts: usage 59, staying 63, problems 9, upkeep 100

    PyPI · GitHub

60. 60agent-framework-coreAI and Language ModelsVendor-backedDownloads / month1,385,734Retention kept after 3, 6 months66% / 100%HoldingChurn lost after 3, 6 months34% / 0%ProblemsHighScore / 10059

    Microsoft Agent Framework for building AI Agents with Python. This is the core package that has all the core abstractions and implementations.

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 66% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 34% / 0%
    • Change vs 6 months ago: +76%
    • GitHub issues opened in the last 12 months: 3,129
    • GitHub stars: 14,384
    • Last release: 2026-10-22, last code push 2026-10-30
    • Score parts: usage 47, staying 83, problems 0, upkeep 100

    PyPI · GitHub

61. 61langchain-litellmAI and Language ModelsVendor-backedDownloads / month1,334,368Retention kept after 3, 6 months37% / 100%FadingChurn lost after 3, 6 months63% / 0%ProblemsHighScore / 10059

    An integration package connecting LiteLLM and LangChain

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 37% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 63% / 0%
    • Change vs 6 months ago: +16%
    • GitHub issues opened in the last 12 months: 75
    • GitHub stars: 61
    • Last release: 2026-10-25, last code push 2026-10-29
    • Score parts: usage 47, staying 69, problems 31, upkeep 100

    PyPI · GitHub

62. 62fireworks-aiAI and Language ModelsDownloads / month1,500,156Retention kept after 3, 6 months38% / 27%FadingChurn lost after 3, 6 months62% / 73%ProblemsLowScore / 10058

    The official Python library for the fireworks API

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 38% of the download level
    • Still there after 6 months: 27%
    • Download churn (level lost) after 3 / 6 months: 62% / 73%
    • Change vs 6 months ago: −73%
    • GitHub issues opened in the last 12 months: 3
    • GitHub stars: 11
    • Last release: 2026-10-21, last code push 2026-10-30
    • Score parts: usage 48, staying 32, problems 91, upkeep 100

    PyPI · GitHub

63. 63pydantic-aiAI and Language Models · +1 related packagesVendor-backedDownloads / month6,832,854Retention kept after 3, 6 months44% / 28%FadingChurn lost after 3, 6 months56% / 72%ProblemsMediumScore / 10057

    AI Agent Framework, the Pydantic way

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 44% of the download level
    • Still there after 6 months: 28%
    • Download churn (level lost) after 3 / 6 months: 56% / 72%
    • Change vs 6 months ago: −72%
    • GitHub issues opened in the last 12 months: 3,039 (shared by the whole repo)
    • GitHub stars: 20,952
    • Last release: 2026-10-28, last code push 2026-10-30
    • Same project, not listed separately: pydantic-ai-harness
    • Score parts: usage 63, staying 36, problems 45, upkeep 100

    PyPI · GitHub

64. 64ai21AI and Language ModelsDownloads / month354,592Retention kept after 3, 6 months100% / 100%StickyChurn lost after 3, 6 months0% / 0%ProblemsNo repo foundScore / 10057

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 100% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 0% / 0%
    • Change vs 6 months ago: +212%
    • GitHub issues opened in the last 12 months: no repo found
    • GitHub stars: none found
    • Last release: 2025-11-06
    • Score parts: usage 34, staying 100, problems 50, upkeep 40

    PyPI

65. 65opikDevOpsVendor-backedDownloads / month2,281,818Retention kept after 3, 6 months65% / 47%FadingChurn lost after 3, 6 months35% / 53%ProblemsHighScore / 10055

    Comet tool for logging and evaluating LLM traces

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 65% of the download level
    • Still there after 6 months: 47%
    • Download churn (level lost) after 3 / 6 months: 35% / 53%
    • Change vs 6 months ago: −53%
    • GitHub issues opened in the last 12 months: 332
    • GitHub stars: 23,033
    • Last release: 2026-10-30, last code push 2026-10-30
    • Score parts: usage 52, staying 56, problems 18, upkeep 100

    PyPI · GitHub

66. 66mistral-vibeAI and Language ModelsVendor-backedDownloads / month1,580,558Retention kept after 3, 6 months21% / 100%FadingChurn lost after 3, 6 months79% / 0%ProblemsHighScore / 10055

    Minimal CLI coding agent by Mistral

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 21% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 79% / 0%
    • Change vs 6 months ago: +144%
    • GitHub issues opened in the last 12 months: 643
    • GitHub stars: 5,239
    • Last release: 2026-10-26, last code push 2026-10-29
    • Score parts: usage 48, staying 60, problems 15, upkeep 100

    PyPI · GitHub

67. 67weaviate-clientData and StorageVendor-backedDownloads / month4,900,546Retention kept after 3, 6 months2% / 5%FadingChurn lost after 3, 6 months98% / 95%ProblemsLowScore / 10054

    A python native Weaviate client

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 2% of the download level
    • Still there after 6 months: 5%
    • Download churn (level lost) after 3 / 6 months: 98% / 95%
    • Change vs 6 months ago: −95%
    • GitHub issues opened in the last 12 months: 67
    • GitHub stars: 229
    • Last release: 2026-10-26, last code push 2026-10-28
    • Score parts: usage 59, staying 4, problems 92, upkeep 100

    PyPI · GitHub

68. 68doclingData and StorageVendor-backedDownloads / month3,878,252Retention kept after 3, 6 months23% / 68%FadingChurn lost after 3, 6 months77% / 32%ProblemsHighScore / 10054

    SDK and CLI for parsing PDF, DOCX, HTML, and more, to a unified document representation for powering downstream workflows such as gen AI applications.

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 23% of the download level
    • Still there after 6 months: 68%
    • Download churn (level lost) after 3 / 6 months: 77% / 32%
    • Change vs 6 months ago: −32%
    • GitHub issues opened in the last 12 months: 952
    • GitHub stars: 69,850
    • Last release: 2026-10-29, last code push 2026-10-30
    • Score parts: usage 57, staying 45, problems 20, upkeep 100

    PyPI · GitHub

69. 69pinecone-clientData and StorageDownloads / month1,195,586Retention kept after 3, 6 months100% / 73%HoldingChurn lost after 3, 6 months0% / 27%ProblemsNo repo foundScore / 10053

    Pinecone client (DEPRECATED)

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 100% of the download level
    • Still there after 6 months: 73%
    • Download churn (level lost) after 3 / 6 months: 0% / 27%
    • Change vs 6 months ago: −27%
    • GitHub issues opened in the last 12 months: no repo found
    • GitHub stars: none found
    • Last release: 2025-02-21
    • Score parts: usage 46, staying 86, problems 50, upkeep 10

    PyPI

70. 70ragasDevOpsDownloads / month1,343,272Retention kept after 3, 6 months79% / 92%HoldingChurn lost after 3, 6 months21% / 8%ProblemsHighScore / 10052

    Evaluation framework for RAG and LLM applications

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 79% of the download level
    • Still there after 6 months: 92%
    • Download churn (level lost) after 3 / 6 months: 21% / 8%
    • Change vs 6 months ago: −8%
    • GitHub issues opened in the last 12 months: 198
    • GitHub stars: 16,188
    • Last release: 2026-01-13, last code push 2026-02-24
    • Score parts: usage 47, staying 85, problems 12, upkeep 40

    PyPI · GitHub

71. 71crewaiAI and Language Models · +2 related packagesVendor-backedDownloads / month2,600,976Retention kept after 3, 6 months26% / 37%FadingChurn lost after 3, 6 months74% / 63%ProblemsHighScore / 10050

    Cutting-edge framework for orchestrating role-playing, autonomous AI agents. By fostering collaborative intelligence, CrewAI empowers agents to work together seamlessly, tackling complex tasks.

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 26% of the download level
    • Still there after 6 months: 37%
    • Download churn (level lost) after 3 / 6 months: 74% / 63%
    • Change vs 6 months ago: −63%
    • GitHub issues opened in the last 12 months: 1,145 (shared by the whole repo)
    • GitHub stars: 60,285
    • Last release: 2026-10-30, last code push 2026-10-30
    • Same project, not listed separately: crewai-cli, crewai-tools
    • Score parts: usage 53, staying 32, problems 29, upkeep 100

    PyPI · GitHub

72. 72fastapi-mcpWeb and APIsDownloads / month3,992,828Retention kept after 3, 6 months37% / 23%FadingChurn lost after 3, 6 months63% / 77%ProblemsLowScore / 10049

    Automatic MCP server generator for FastAPI applications - converts FastAPI endpoints to MCP tools for LLM integration

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 37% of the download level
    • Still there after 6 months: 23%
    • Download churn (level lost) after 3 / 6 months: 63% / 77%
    • Change vs 6 months ago: −77%
    • GitHub issues opened in the last 12 months: 35
    • GitHub stars: 12,024
    • Last release: 2025-07-28, last code push 2025-11-24
    • Score parts: usage 57, staying 30, problems 72, upkeep 40

    PyPI · GitHub

73. 73mcpoWeb and APIsSudden jumpDownloads / month1,631,112Retention kept after 3, 6 months100% / 100%StickyChurn lost after 3, 6 months0% / 0%ProblemsNo repo foundScore / 10047

    A simple, secure MCP-to-OpenAPI proxy server

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 100% of the download level
    • Still there after 6 months: 100%
    • Download churn (level lost) after 3 / 6 months: 0% / 0%
    • Change vs 6 months ago: +780%
    • GitHub issues opened in the last 12 months: no repo found
    • GitHub stars: none found
    • Last release: 2026-02-27
    • Score parts: usage 49, staying 100, problems 50, upkeep 40, sudden-jump penalty ×0.75

    PyPI

74. 74vllmAI and Language ModelsVendor-backedDownloads / month1,914,514Retention kept after 3, 6 months34% / 29%FadingChurn lost after 3, 6 months66% / 71%ProblemsHighScore / 10045

    A high-throughput and memory-efficient inference and serving engine for LLMs

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 34% of the download level
    • Still there after 6 months: 29%
    • Download churn (level lost) after 3 / 6 months: 66% / 71%
    • Change vs 6 months ago: −71%
    • GitHub issues opened in the last 12 months: 7,891
    • GitHub stars: 95,899
    • Last release: 2026-10-18, last code push 2026-10-30
    • Score parts: usage 50, staying 32, problems 2, upkeep 100

    PyPI · GitHub

75. 75llama-parseData and StorageDownloads / month2,184,266Retention kept after 3, 6 months3% / 9%FadingChurn lost after 3, 6 months97% / 91%ProblemsNo repo foundScore / 10036

    Parse files into RAG-Optimized formats.

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 3% of the download level
    • Still there after 6 months: 9%
    • Download churn (level lost) after 3 / 6 months: 97% / 91%
    • Change vs 6 months ago: −91%
    • GitHub issues opened in the last 12 months: no repo found
    • GitHub stars: none found
    • Last release: 2026-02-13
    • Score parts: usage 52, staying 6, problems 50, upkeep 40

    PyPI

76. 76llama-cloud-servicesData and StorageDownloads / month1,644,356Retention kept after 3, 6 months2% / 7%FadingChurn lost after 3, 6 months98% / 93%ProblemsNo repo foundScore / 10034

    Tailored SDK clients for LlamaCloud services.

    Loading chart…

    Loading chart…

    Typical daily downloads (median day) in 28-day blocks; each bar is the share of the previous block's level lost. This is download churn, not user churn.

    • Still there after 3 months: 2% of the download level
    • Still there after 6 months: 7%
    • Download churn (level lost) after 3 / 6 months: 98% / 93%
    • Change vs 6 months ago: −93%
    • GitHub issues opened in the last 12 months: no repo found
    • GitHub stars: none found
    • Last release: 2026-02-13
    • Score parts: usage 49, staying 5, problems 50, upkeep 40

    PyPI

No packages match. Try another word or pick "All".

Downloads / month = median day in the last 28 days × 28. Kept after 3 / 6 months (retention) = today's download level as a share of the level 3 and 6 months ago, capped at 100%. The 6-month figure for PyPI uses a shorter window, see the methodology. Lost after 3 / 6 months (churn) = 100% minus that. Both are about downloads, not users, and growth shows as 0% lost. Sticky = kept at least 90% after 3 and 6 months; Holding = 60 to 90%; Fading = under 60%. Problems: GitHub issues opened in the last 12 months per 100,000 downloads, Low/Medium/High against this list. Vendor-backed = the GitHub repo belongs to the company behind the model or product. Sudden jump = downloads up over 6× in 6 months with little GitHub activity; its score is cut by a quarter.

New and rising (under 6 months old)

Too new to measure staying power, so not scored.

• openai-codexopenai-codex · AI and Language Models2,086,336 downloads / month · first published 2026-05-28

Packages we leave out

These clear the download bar but mostly arrive as a dependency of another package, so their counts say little about how many people chose them. The list is kept by hand.

• huggingface-hub: installed by transformers, tokenizers and datasets
• tokenizers: installed by transformers and sentence-transformers
• mcp-types: installed by mcp
• langsmith: installed by langchain-core
• langchain-protocol: installed by LangGraph
• langgraph-sdk: installed by langgraph
• fastmcp-slim: installed by fastmcp
• langgraph-checkpoint: installed by langgraph
• langgraph-prebuilt: installed by langgraph
• langchain-text-splitters: installed by langchain
• genai-prices: installed by pydantic-ai
• pydantic-ai-slim: installed by pydantic-ai
• openinference-semantic-conventions: installed by OpenInference instrumentors
• openinference-instrumentation: installed by OpenInference instrumentors
• openai-harmony: installed by vLLM
• litellm-proxy-extras: installed by litellm[proxy]
• litellm-enterprise: installed by litellm[proxy]
• docling-core: installed by docling
• llama-index-workflows: installed by llama-index-core
• llama-index-instrumentation: installed by llama-index-core
• mistral-common: installed by vLLM
• opentelemetry-util-genai: installed by OpenTelemetry GenAI instrumentors
• llguidance: installed by vLLM and other inference engines
• docling-parse: installed by docling
• docling-ibm-models: installed by docling
• outlines-core: installed by outlines and vLLM
• docling-slim: installed by docling
• pinecone-plugin-interface: installed by pinecone
• llama-cloud: installed by llama-index packages
• langgraph-runtime-inmem: installed by langgraph-cli
• langgraph-api: installed by langgraph-cli
• openai-codex-cli-bin: installed by openai-codex
• crewai-core: installed by crewai
• pinecone-plugin-assistant: installed by pinecone
• microsoft-agents-activity: installed by the Microsoft 365 Agents SDK
• microsoft-agents-hosting-core: installed by the Microsoft 365 Agents SDK
• loongsuite-util-genai: installed by LoongSuite instrumentors

Related

• All rankingsn8n nodes, npm and Python
• AI npm packagesJavaScript and TypeScript
• n8n community nodesThe same method for n8n nodes
• MethodologySources, formula and limits

Raw data: data.json. Sources: pypistats.org and the PyPI JSON API, GitHub.