Best Agent frameworks — Opt out of training

The ranking

1. ★ Editor’s pick

   #1LA

   LangGraph

   Open-source orchestration framework and runtime for stateful, long-running LLM agents, from LangChain.

   Best forEngineering teams building production agents who need cyclical graphs, durable state and replayable runs

   Opt out of training: Yes — Nothing to opt out of: the vendor states it does not train on customer data. Separately, setting LANGSMITH_TRACING=false keeps agent data on your own machine entirely.“No. LangSmith does not use your data to train models. Your traces, prompts, and outputs remain private to your organization.”General Questions FAQ, 'Will you train on the data that I send to LangSmith?' · retrieved 2026-09-02

   Visit LangGraph ↗Read the review →Compare LangGraphGet a quote →From Free
2. #2AG

   Agno

   Python SDK, open-source AgentOS runtime and hosted control plane for building and operating agents.

   Best forPython developers who want an Apache 2.0 SDK with memory, knowledge, learning, guardrails and a Control Plane they run inside their own cloud account, and who are comfortable running Docker or a cloud deploy template themselves.

   Opt out of training: Yes — Nothing to opt out of: the vendor never receives a copy of your data, which stays in your own database.“Your data stays in your cloud and we never hold a copy of it.”Plan comparison table, Control plane section footnote (final sentence) · retrieved 2026-09-02

   Visit Agno ↗Read the review →Compare AgnoGet a quote →From Free
3. #3CS

   Claude Agent SDK

   Python and TypeScript library that runs Claude Code's agent loop and tools inside your own application.

   Best forDevelopers already building on Claude (or Claude Code) who want its exact agent loop, tools and context management as a callable library, and are comfortable being single-vendor on Anthropic in exchange for the tightest possible integration with Claude Code's own capabilities.

   Opt out of training: Yes — Retained API data is not used for training without express permission, and the opt-in route is a named program an admin must join.“Retained data is never used for model training without your express permission.”How Anthropic approaches data retention, first commitment bullet · retrieved 2026-09-02

   Visit Claude Agent SDK ↗Read the review →Compare Claude Agent SDKGet a quote →From $1/million input tokens

4. Advertisement

   Your agent here. Go Featured and sit above the free listings in this ranking, or buy the slot beside it. Every paid unit and every paid row wears the Advertisement label above — and the verdict, the facts, and what we leave out are not for sale.

   Get featured · top of your category →
5. #4MA

   Mastra

   Open-source TypeScript framework for building AI agents, with a hosted platform for running and observing them.

   Best forFull-stack JavaScript and TypeScript teams, especially those already building in Next.js or Node, who want agents, durable workflows and memory unified in one framework without adopting a Python-first stack.

   Opt out of training: Yes — Contractual no-training default on customer inputs and outputs, for Mastra's and model providers' models, with training only on express opt-in.“Mastra does not, and will not, use Customer Inputs or Customer Outputs to train, fine-tune, or otherwise improve any foundation model or other Model owned by Mastra or by any Model Provider”Section 4.3 No Training on Customer Data · retrieved 2026-09-02

   Visit Mastra ↗Read the review →Compare MastraGet a quote →From Free
6. #5LA

   LiveKit Agents

   Open-source Python and Node.js framework for voice AI agents, with LiveKit Cloud hosting, telephony and observability.

   Best forEngineering teams that want to own their voice-agent stack end to end, self-hosted or on LiveKit Cloud, without being locked into one vendor's API.

   Opt out of training: Yes — Nothing to opt out of: zero retention is the default on every plan, LiveKit logs no prompts, audio or completions, and voice-clone samples are opted out of provider training on your behalf.“Sub-processors do not retain, log, or use customer inference traffic for training or model improvement.”AI and inference data section, 'Zero retention by sub-processors' · retrieved 2026-09-02

   Visit LiveKit Agents ↗Read the review →Compare LiveKit AgentsGet a quote →From Free
7. #6LL

   LlamaIndex

   Open-source agent framework plus a credit-metered document parsing, extraction and indexing platform.

   Best forEngineers building RAG-heavy or document-heavy agents who want an MIT-licensed OSS core with an optional managed parsing and extraction platform on top.

   Opt out of training: Yes — Nothing is used for training by default, and do_not_cache=True skips the 48-hour cache entirely.“Your data is kept private to you only and is used only to return your results, never for model training.”Privacy & Caching, Are documents stored or cached? · retrieved 2026-09-02

   Visit LlamaIndex ↗Read the review →Compare LlamaIndexGet a quote →From Free
8. #7LA

   LangChain

   Open source agent frameworks plus LangSmith, a hosted platform to trace, evaluate, deploy and govern agents.

   Best forEngineering teams building a bespoke, production-grade agent who want the largest integration ecosystem plus real stateful control.

   Opt out of training: Yes — No opt-out is needed because the vendor states it never trains models on customer traces, prompts or outputs; Engine's analysis still runs through LangChain-managed inference.“No. LangSmith does not use your data to train models. Your traces, prompts, and outputs remain private to your organization.”FAQ, General Questions, 'Will you train on the data that I send to LangSmith?' · retrieved 2026-09-18

   Visit LangChain ↗Read the review →Compare LangChainGet a quote →From Free