Sharing Agno, a new open-source library focused on building high-performance, multimodal AI agents. If you're building agentic systems, this looks seriously impressive, especially regarding speed and memory efficiency.
Agno acts as a lightweight framework, providing a unified API for various LLMs and adding capabilities like memory, knowledge stores, tool use, and reasoning.
Key aspects that stand out:
🚀 Lightning Fast & Lightweight: They report huge performance gains over frameworks like LangGraph (claiming 10,000x faster instantiation and 50x less memory on their benchmarks). 🔌 Model Agnostic: No lock-in! Use models from OpenAI, Anthropic, Cohere, or open-source ones via Ollama, Together, Anyscale, etc. 👁️ Multimodal: Native support for agents working with text, image, audio, and video. 🤝 Multi-Agent Teams: Built to orchestrate teams of specialized agents. 🧠 Memory, Knowledge, Tools: Built-in support for memory, vector DBs (for RAG), and adding custom tools. 📊 Monitoring: Integrates with agno.com for real-time agent monitoring. 🔓 Open Source (Apache 2.0): Freely available for use and contribution.
For developers building high-performance, multimodal AI agents, Agno offers a powerful and efficient open-source foundation.
@sentry_co@zaczuo Hey Zac, Nice work. Congrats. btw, what do you provide for Audio and Video AI agents as LLM options. Not every models excels at these.
@ansub Congrats on launching Agno! Building a lightweight and open-source library for multimodal AI agents is a fantastic contribution to the community.
Agno looks super promising—love the clean approach to collaborative knowledge management. Having a single, organized space to keep the team aligned and reduce information chaos is exactly what teams (and brains!) need. Definitely bookmarking it to test with my workflow 🙏
@ansub Agno’s focus on speed, multimodal flexibility, and open-source ethos makes it a tantalizing foundation for developers pushing agentic systems beyond today’s latency-heavy norms. The claims around memory efficiency and multi-agent orchestration could redefine how teams scale AI workflows. But how does Agno handle resource contention in complex, real-world deployments—say, when 100+ agents with competing priorities access shared tools or memory? Does the framework prioritize tasks dynamically (e.g., via cost-based optimization), or does it rely on developers to manually define agent hierarchies?
Fantastic and truly promising! Maybe we should partner up, using agents in a fintech data crunching context. Even being multimodal comes into a vital play for us. Thoughts? @Agno
Replies
Hi everyone!
Sharing Agno, a new open-source library focused on building high-performance, multimodal AI agents. If you're building agentic systems, this looks seriously impressive, especially regarding speed and memory efficiency.
Agno acts as a lightweight framework, providing a unified API for various LLMs and adding capabilities like memory, knowledge stores, tool use, and reasoning.
Key aspects that stand out:
🚀 Lightning Fast & Lightweight: They report huge performance gains over frameworks like LangGraph (claiming 10,000x faster instantiation and 50x less memory on their benchmarks).
🔌 Model Agnostic: No lock-in! Use models from OpenAI, Anthropic, Cohere, or open-source ones via Ollama, Together, Anyscale, etc.
👁️ Multimodal: Native support for agents working with text, image, audio, and video.
🤝 Multi-Agent Teams: Built to orchestrate teams of specialized agents.
🧠 Memory, Knowledge, Tools: Built-in support for memory, vector DBs (for RAG), and adding custom tools.
📊 Monitoring: Integrates with agno.com for real-time agent monitoring.
🔓 Open Source (Apache 2.0): Freely available for use and contribution.
For developers building high-performance, multimodal AI agents, Agno offers a powerful and efficient open-source foundation.
Hunting credit to @sentry_co 🙌
@sentry_co @zaczuo Hey Zac, Nice work. Congrats. btw, what do you provide for Audio and Video AI agents as LLM options. Not every models excels at these.
@sentry_co @zaczuo @imraju Hey!
Agno is model-agnostic, so you can choose the best models for your use case.
We’ve actually documented a few multimodal use cases here:
docs.agno.com/examples/concepts/multimodal
Let us know what you're building—would love to help brainstorm the right setup!
@sentry_co @zaczuo Beautiful design. Thanks for hunting.
@ansub Congrats on launching Agno! Building a lightweight and open-source library for multimodal AI agents is a fantastic contribution to the community.
Agno looks super promising—love the clean approach to collaborative knowledge management. Having a single, organized space to keep the team aligned and reduce information chaos is exactly what teams (and brains!) need. Definitely bookmarking it to test with my workflow 🙏
This looks cool! Coming to typescript any time soon? :)
So good, best of luck guys!
Wing Assistant
I built a multi-agent system with Agno, really simple to use. Great job!
Model-agnostic agents simplify my workflow! Thanks! 👍
chatWise
Hey there! Tried it out and loved the one time link feature. Super useful when I don’t want the message hanging around.
https://www.linkedin.com/posts/thisisnish_ailearning-crewai-flow-activity-7315521650052026368-K05D
Check out my review after building AI agents using Agno and CrewAI. IMO, Agno FTW!
Super cool! Congrats on the launch
I love Agno. It's so simple to use even I with my low dev knowledge can build powerful agents.
I built an RSA generator and working on my own Google Ads agent using Agno.
Great job team.
@ansub Agno’s focus on speed, multimodal flexibility, and open-source ethos makes it a tantalizing foundation for developers pushing agentic systems beyond today’s latency-heavy norms. The claims around memory efficiency and multi-agent orchestration could redefine how teams scale AI workflows. But how does Agno handle resource contention in complex, real-world deployments—say, when 100+ agents with competing priorities access shared tools or memory? Does the framework prioritize tasks dynamically (e.g., via cost-based optimization), or does it rely on developers to manually define agent hierarchies?
Best AI framework.
Fantastic and truly promising! Maybe we should partner up, using agents in a fintech data crunching context. Even being multimodal comes into a vital play for us. Thoughts? @Agno