LocalAIAgents · 548 frameworks
LocalAIAgents ranks the best AI agent frameworks on real GitHub activity: 548 frameworks scored on stars, forks and maintenance, with LangGraph vs CrewAI vs AutoGen verdicts and local-LLM support flags. I also build custom AI agents for businesses, scoped to your actual workflow.
Sample preview of the agent framework database.
●Direct answers
The questions everyone asks, answered straight.
It depends on how much control you want. LangGraph gives Python developers fine-grained control over agent workflows. CrewAI is friendlier for role-based multi-agent teams. AutoGen pioneered conversational agent patterns. Our verdicts compare the best open source AI agent frameworks on activity, license and capabilities. Pick the one that matches your project, not the hype.
One with a pluggable model layer, not locked to a single provider’s API. Plenty of frameworks now support Ollama and other local runtimes out of the box. We flag which frameworks officially support local LLMs and self-hosted agent runtimes before you commit.
Python has the mature options; TypeScript fits JavaScript stacks. Python dominates the agent ecosystem, while Mastra and the Vercel AI SDK suit teams building in JS. We separate frameworks by language, with stars, activity and license for each, so the choice is obvious.
●Features
Real GitHub signals, head-to-head verdicts, local-LLM support. All in one place.
Stars, forks, contributors, maintenance signals. Compiled from GitHub across hundreds of agent frameworks, refreshed from real data.
LangGraph vs CrewAI vs AutoGen vs OpenAgents, compared on activity, capabilities and license. Opinions included.
Python staples on one side, Mastra and the Vercel AI SDK on the other. Filter by the language you actually write.
Which frameworks run on local LLMs and self-hosted runtimes, and which are genuinely production-ready.
●The dataset
What we collected, field by field. Every signal traceable to its repo.
| Ecosystem | Frameworks | Key fields |
|---|---|---|
| Python frameworks | 250+ | Stars, forks, maintenance signals |
| TypeScript frameworks | 100+ | Activity, license, capabilities |
| Multi-agent orchestration | 120+ | Orchestration style, language |
| Runtimes & tooling | 70+ | Local-LLM support, self-hosting |
| Field | What it means | Sample value |
|---|---|---|
| GitHub stars | Popularity signal | 12.4K |
| Last commit | Maintenance freshness | 3 days ago |
| License | Usage terms | MIT |
| Local-LLM support | Works with local runtimes | yes |
●Preview
A peek under the hood. This is the data I work from.
●Topic hub
One keyword would not do it justice. The whole AI agent framework topic, mapped:
All 548 frameworks, ranked on stars, forks and maintenance.
LangGraph vs CrewAI vs AutoGen and more, with actual verdicts.
The most mature agent ecosystem, measured by the numbers.
Mastra, the Vercel AI SDK, and the JS side of the agent world.
The best AI agent framework for local LLMs, flagged per framework.
The maintenance signals that separate weekend experiments from real tooling.
Rankings built on GitHub data, not marketing copy.
Dozens of AI agent frameworks launched in the last two years, and their marketing pages all claim the same things. LocalAIAgents’ AI agent framework comparison ranks the best AI agent frameworks on GitHub data: stars, forks, contributors, maintenance. So you can see which open source AI agent frameworks are actually alive.
The big matchups matter. LangGraph vs CrewAI vs AutoGen is the question most teams ask first, so we publish head-to-head verdicts. We also track CrewAI alternatives and LangChain alternatives, and compare multi-agent frameworks on orchestration style, language and license.
Pick by your stack. The best Python AI agent framework if you build in Python, TypeScript options if you don’t. Hunting for the best open source AI agent builder projects or the top AI agent projects on GitHub? Our rankings update from real activity data. Build my agent to put this data to work for you.
●FAQ
Answers about picking frameworks, languages, and local-LLM support
●References
Hugging Face: Open-weight models that power many local agents, all in one hub.
arXiv: The papers behind agent architectures and methods.
GitHub: Repos, stars and activity data. Where our rankings come from.
Ollama: Local models your agent frameworks can call, running on your machine.
●Work with me
Describe the job. I build the agent around it, on the framework that fits.
$799 starting price
Tell me about your situation and I will get back to you.