AgentScout Logo Agent Scout

NPM AI Packages Download Tracker — Week of May 3, 2026

Mistral AI SDK surged 58.9% to 3.7M weekly downloads while Anthropic SDK grew 22.2% to 17.8M, narrowing the gap with OpenAI SDK (20.9M). Weekly tracker of AI/LLM package adoption on NPM.

AgentScout · · · 4 min read
#npm #ai-packages #downloads #sdk #framework #tracker
Analyzing Data Nodes...
SIG_CONF:CALCULATING
Verified Sources

Data Overview

  • Snapshot Week: 2026-04-25 to 2026-05-01
  • Tracker: NPM AI Packages Download Tracker (view all snapshots: /tech/ai-agents/data/?tracker=npm-ai-packages-tracker)
  • Update Frequency: Weekly
  • Primary Sources: NPM Downloads API, NPM Registry Search API

Key Facts

  • Who: 18 AI/LLM packages tracked across SDKs, frameworks, and vector databases
  • What: 81.4M total weekly downloads tracked; Mistral AI SDK recorded the largest growth (+58.9%)
  • When: Week of May 3, 2026 (data collection period: April 25 - May 1, 2026)
  • Impact: 3 high-significance changes identified, including emergence of Mastra framework

Methodology

Data collected via NPM Downloads API and NPM Registry Search API. Weekly download counts represent the total number of installations from npm registry over the 7-day period ending May 1, 2026. Packages are categorized as SDK (official client libraries), Framework (development frameworks), or VectorDB (vector database clients). Version information reflects the latest published version at time of collection.

Inclusion Criteria: Packages are selected based on relevance to AI/LLM development, minimum download threshold (30K weekly), and active maintenance status. New packages meeting criteria are added upon discovery.

This Week’s Data

PackageWeekly DownloadsVersionCategoryDescription
openai20,962,4616.35.0SDKThe official TypeScript library for the OpenAI API
@anthropic-ai/sdk17,842,5700.92.0SDKThe official TypeScript library for the Anthropic API
ai12,078,9956.0.174FrameworkVercel AI SDK - Build AI-powered applications with React, Svelte, Vue, and Solid
@ai-sdk/openai5,942,6893.0.58FrameworkOpenAI provider for Vercel AI SDK
@ai-sdk/anthropic5,320,6503.0.74FrameworkAnthropic provider for Vercel AI SDK
@anthropic-ai/claude-agent-sdk4,927,8630.2.126FrameworkAnthropic’s official agent SDK for Claude
@langchain/core3,845,7241.1.44FrameworkCore LangChain.js abstractions and schemas
@mistralai/mistralai3,711,8022.2.1SDKTypeScript client library for the Mistral AI API
@langchain/openai2,877,7661.4.5FrameworkOpenAI integrations for LangChain.js
@google/generative-ai2,854,9990.24.1SDKGoogle Generative AI SDK for JavaScript
langchain2,052,3301.3.5FrameworkLangChain framework for building LLM applications
@elastic/elasticsearch2,015,937latestVectorDBElasticsearch client for Node.js
@mastra/core795,4071.31.0FrameworkMastra - Framework for building AI-powered applications and agents
@pinecone-database/pinecone593,2757.2.0VectorDBPinecone vector database client
cohere-ai414,7828.0.0SDKCohere TypeScript SDK
llamaindex111,1430.12.1FrameworkLlamaIndex TypeScript - Data framework for LLM applications
vectra35,4010.14.0VectorDBLightweight local vector database for Node.js
@langchain/anthropic731,397latestFrameworkAnthropic integrations for LangChain.js

Week-over-Week Summary

MetricThis WeekLast WeekChange
Total tracked downloads81,450,15779,475,891+1,974,266 (+2.5%)
Packages tracked1810+8 (new entries)
SDK category total44,807,01438,289,535+6,517,479 (+17.0%)
Framework category total33,955,41530,995,121+2,960,294 (+9.5%)
VectorDB category total2,644,713N/Anew category
Highest growth@mistralai/mistralai+58.9%+1,375,137 downloads
  • Mistral AI SDK explosive growth: @mistralai/mistralai weekly downloads surged 58.9% (from 2.3M to 3.7M), marking the largest single-week growth among major AI SDKs. This signals rapid adoption of European LLM alternatives, particularly in developer communities seeking OpenAI alternatives.

  • Anthropic SDK closing gap with OpenAI: The download gap between OpenAI SDK (20.9M) and Anthropic SDK (17.8M) narrowed from 3.9M to 3.1M weekly installs. Anthropic’s 22.2% growth rate significantly outpaced OpenAI’s 13.1%, suggesting accelerating Claude adoption.

  • Vercel AI SDK ecosystem stability: The @ai-sdk/* packages maintained steady 5-7% growth, indicating mature adoption patterns. Combined Vercel AI ecosystem now totals 23.3M weekly downloads across all providers.

  • Mastra framework emergence: @mastra/core appeared in tracking with 795K weekly downloads, establishing itself as a notable new contender in the agent framework space alongside LangChain and Vercel AI SDK.

Category Breakdown

CategoryTotal DownloadsAvg GrowthKey Players
SDK44,807,014+18.5%openai, @anthropic-ai/sdk, @mistralai/mistralai, @google/generative-ai, cohere-ai
Framework33,955,415+4.8%ai, @ai-sdk/*, langchain, @mastra/core, llamaindex
VectorDB2,644,713N/A@elastic/elasticsearch, @pinecone-database/pinecone, vectra

Notable Changes This Week

EventDescriptionSignificance
Mistral AI SDK explosive growth@mistralai/mistralai weekly downloads surged 58.9% (from 2.3M to 3.7M), marking the largest single-week growth among major AI SDKsHigh
Anthropic SDK closing gap with OpenAIAnthropic SDK grew 22.2% while OpenAI grew 13.1%, narrowing the download gap from 3.9M to 3.1M weekly installsMedium
Mastra framework emergence@mastra/core appeared in tracking with 795K weekly downloads, establishing itself as a notable new agent frameworkMedium

🔺 Scout Intel: What Others Missed

Confidence: high | Novelty Score: 65/100

While download counts are publicly available, the strategic signal lies in the growth differentials. Mistral’s 58.9% surge — the largest among major SDKs — indicates developers are actively diversifying away from OpenAI dependency. Combined with Anthropic’s 22.2% growth, non-OpenAI SDKs captured 73% of this week’s net new installs (2.1M of 2.9M total growth). This represents a measurable shift in developer sentiment toward multi-model architectures, a trend not visible in individual package metrics but clear when analyzing the ecosystem holistically.

Key Implication: Infrastructure teams should prioritize multi-provider SDK abstractions now rather than later — the developer community is already voting with their package managers, and single-provider architectures will face increasing technical debt as multi-model patterns become the default.

Sources

NPM AI Packages Download Tracker — Week of May 3, 2026

Mistral AI SDK surged 58.9% to 3.7M weekly downloads while Anthropic SDK grew 22.2% to 17.8M, narrowing the gap with OpenAI SDK (20.9M). Weekly tracker of AI/LLM package adoption on NPM.

AgentScout · · · 4 min read
#npm #ai-packages #downloads #sdk #framework #tracker
Analyzing Data Nodes...
SIG_CONF:CALCULATING
Verified Sources

Data Overview

  • Snapshot Week: 2026-04-25 to 2026-05-01
  • Tracker: NPM AI Packages Download Tracker (view all snapshots: /tech/ai-agents/data/?tracker=npm-ai-packages-tracker)
  • Update Frequency: Weekly
  • Primary Sources: NPM Downloads API, NPM Registry Search API

Key Facts

  • Who: 18 AI/LLM packages tracked across SDKs, frameworks, and vector databases
  • What: 81.4M total weekly downloads tracked; Mistral AI SDK recorded the largest growth (+58.9%)
  • When: Week of May 3, 2026 (data collection period: April 25 - May 1, 2026)
  • Impact: 3 high-significance changes identified, including emergence of Mastra framework

Methodology

Data collected via NPM Downloads API and NPM Registry Search API. Weekly download counts represent the total number of installations from npm registry over the 7-day period ending May 1, 2026. Packages are categorized as SDK (official client libraries), Framework (development frameworks), or VectorDB (vector database clients). Version information reflects the latest published version at time of collection.

Inclusion Criteria: Packages are selected based on relevance to AI/LLM development, minimum download threshold (30K weekly), and active maintenance status. New packages meeting criteria are added upon discovery.

This Week’s Data

PackageWeekly DownloadsVersionCategoryDescription
openai20,962,4616.35.0SDKThe official TypeScript library for the OpenAI API
@anthropic-ai/sdk17,842,5700.92.0SDKThe official TypeScript library for the Anthropic API
ai12,078,9956.0.174FrameworkVercel AI SDK - Build AI-powered applications with React, Svelte, Vue, and Solid
@ai-sdk/openai5,942,6893.0.58FrameworkOpenAI provider for Vercel AI SDK
@ai-sdk/anthropic5,320,6503.0.74FrameworkAnthropic provider for Vercel AI SDK
@anthropic-ai/claude-agent-sdk4,927,8630.2.126FrameworkAnthropic’s official agent SDK for Claude
@langchain/core3,845,7241.1.44FrameworkCore LangChain.js abstractions and schemas
@mistralai/mistralai3,711,8022.2.1SDKTypeScript client library for the Mistral AI API
@langchain/openai2,877,7661.4.5FrameworkOpenAI integrations for LangChain.js
@google/generative-ai2,854,9990.24.1SDKGoogle Generative AI SDK for JavaScript
langchain2,052,3301.3.5FrameworkLangChain framework for building LLM applications
@elastic/elasticsearch2,015,937latestVectorDBElasticsearch client for Node.js
@mastra/core795,4071.31.0FrameworkMastra - Framework for building AI-powered applications and agents
@pinecone-database/pinecone593,2757.2.0VectorDBPinecone vector database client
cohere-ai414,7828.0.0SDKCohere TypeScript SDK
llamaindex111,1430.12.1FrameworkLlamaIndex TypeScript - Data framework for LLM applications
vectra35,4010.14.0VectorDBLightweight local vector database for Node.js
@langchain/anthropic731,397latestFrameworkAnthropic integrations for LangChain.js

Week-over-Week Summary

MetricThis WeekLast WeekChange
Total tracked downloads81,450,15779,475,891+1,974,266 (+2.5%)
Packages tracked1810+8 (new entries)
SDK category total44,807,01438,289,535+6,517,479 (+17.0%)
Framework category total33,955,41530,995,121+2,960,294 (+9.5%)
VectorDB category total2,644,713N/Anew category
Highest growth@mistralai/mistralai+58.9%+1,375,137 downloads
  • Mistral AI SDK explosive growth: @mistralai/mistralai weekly downloads surged 58.9% (from 2.3M to 3.7M), marking the largest single-week growth among major AI SDKs. This signals rapid adoption of European LLM alternatives, particularly in developer communities seeking OpenAI alternatives.

  • Anthropic SDK closing gap with OpenAI: The download gap between OpenAI SDK (20.9M) and Anthropic SDK (17.8M) narrowed from 3.9M to 3.1M weekly installs. Anthropic’s 22.2% growth rate significantly outpaced OpenAI’s 13.1%, suggesting accelerating Claude adoption.

  • Vercel AI SDK ecosystem stability: The @ai-sdk/* packages maintained steady 5-7% growth, indicating mature adoption patterns. Combined Vercel AI ecosystem now totals 23.3M weekly downloads across all providers.

  • Mastra framework emergence: @mastra/core appeared in tracking with 795K weekly downloads, establishing itself as a notable new contender in the agent framework space alongside LangChain and Vercel AI SDK.

Category Breakdown

CategoryTotal DownloadsAvg GrowthKey Players
SDK44,807,014+18.5%openai, @anthropic-ai/sdk, @mistralai/mistralai, @google/generative-ai, cohere-ai
Framework33,955,415+4.8%ai, @ai-sdk/*, langchain, @mastra/core, llamaindex
VectorDB2,644,713N/A@elastic/elasticsearch, @pinecone-database/pinecone, vectra

Notable Changes This Week

EventDescriptionSignificance
Mistral AI SDK explosive growth@mistralai/mistralai weekly downloads surged 58.9% (from 2.3M to 3.7M), marking the largest single-week growth among major AI SDKsHigh
Anthropic SDK closing gap with OpenAIAnthropic SDK grew 22.2% while OpenAI grew 13.1%, narrowing the download gap from 3.9M to 3.1M weekly installsMedium
Mastra framework emergence@mastra/core appeared in tracking with 795K weekly downloads, establishing itself as a notable new agent frameworkMedium

🔺 Scout Intel: What Others Missed

Confidence: high | Novelty Score: 65/100

While download counts are publicly available, the strategic signal lies in the growth differentials. Mistral’s 58.9% surge — the largest among major SDKs — indicates developers are actively diversifying away from OpenAI dependency. Combined with Anthropic’s 22.2% growth, non-OpenAI SDKs captured 73% of this week’s net new installs (2.1M of 2.9M total growth). This represents a measurable shift in developer sentiment toward multi-model architectures, a trend not visible in individual package metrics but clear when analyzing the ecosystem holistically.

Key Implication: Infrastructure teams should prioritize multi-provider SDK abstractions now rather than later — the developer community is already voting with their package managers, and single-provider architectures will face increasing technical debt as multi-model patterns become the default.

Sources

lhfvhixhwk1s8ap4ga8bw████1mzt7i54uu6j2kl1qvqqyq2p47a67sd9r████3h2g6hcekxlxj1kenpwnwats783fbs8nb████mj2lbkxv66dib9lzkfnvbesbhal58c7c9░░░cz7goe8szmnf8jona5ftkey29502axsf░░░eklqcihqpphwy5fbxe6gl76ie0s9suyc░░░oyku6g07as1ttudfxb4iabdf2udvjvdr░░░kwutljnem2cm08dbc2ymqfle0j9ke35qi░░░pgqa9i7lxw46mxjhmdfvbkbii4a9f3k░░░plo2phla85co9h2i4bgw9t78s26rzzgb████khszkouqo8hylg5nly9k8m8hsi20tc8kv░░░0wuvovsg2utg8qz5844ekzuhd1k29qjk████g7l3jgfy0d7ebqkky0bcnvkodib94mqdr░░░m5ycg853yizqw4k61m9p6icxes7un7a████4o8xod5ewerqfs8a1yxplzr3wvpdtvv░░░lnvl7rx8tusod9elrqi87her67drbd9g████rv39wi8jejrp2imzfp42fz0wiht2h1x████mgbyy5l8pywywj96cq3paupextj19ue░░░g8k5i2l70hr8kmisp909eahjvyadd435g████i4diouuspfpegemh9rzkk4l2ajqha8████wuivunnn59j4ndc9eobx8iv1cckgvsza████6zh738zoy0gs29cz0ir1oualxeetdzii░░░kqypvv4004dwxfv0i97hrm81gtmzgwri████7oliz5podok1ecgcfk3sxftx3i7he7ly████pyk2040whlozz1h1yr01833b0irsyrb6░░░zdva5fp31hoihp7jltpdgo5541qky48e████fcv2mow94974hvtl9nkzwzz4fnrov6d3e████q4xlpwmqt1hzt0xqbmwalug8bbehzxir████f5oxombkgugbdnqib4bhoodno89hokhkj████13p7c2xtpnvio1khvyyln7n3oo1o15pp5m████nawoqn54dawlx5gtqqt4p1rtc1jxhpd████beyr4yznyun0cobnqq0bhg4u8jnnuboq1h████n475ycelxfa7f07yx4uoxyrmxis9e4on░░░kgzz4eddlndj76ged0a3cjts31uscpy2h░░░kj3g8npf87axzde3dtvmyrg6vam0xd4kl████knrl870hf94ags23tj9prmxsdpr4plk░░░cg4id5gqvljf79gu1togx1fkvg09oayp████r0s50yys5rf2d4tja2cl0qshakbkalt2░░░8nx9399q6nm1zjv2tvsm1pljsfxe52dd░░░xokq3559h5otgscgwy3bndbwrrfhrwen8████zftnehm6sucg3etf0lyic51nm3fywowjk████cx8q62kpu5goyd17cvaf4it6nbl9mhdp████p9012z5n5pd2s8a43q54wk5gpyuwe3wxi░░░gmh13igctiqf77uxjs8vtienng672txvb░░░rukzfdhywnit6fnkk5tkgl5l6vkba7rh6████igmbc8zz96qz2hrmujyz9y6x71ar7uut░░░17koo8ciydwx4xixsn5ujo0t2uaa3mnuc████qj6dunqphqsa610q7quzzivyje6ywmokh░░░jqdyluceung49pmgxj3qrsi3kr1ixpbfq████6c7feo3qqomuvxg18rdgei8azr1oyc3o░░░h0z8b74mic