Capability · Framework — agents

Agno

Agno is a lean, Python-native agent framework that emphasises speed — agents instantiate in microseconds and the library has no heavy runtime. It bundles memory, knowledge (built-in vector stores), tools, reasoning, teams for multi-agent collaboration, and a web Playground. It's model-agnostic (OpenAI, Anthropic, Google, Groq, Cohere, Mistral, local) and positions itself as a high-performance LangChain alternative.

Framework facts

Category: agents
Language: Python
License: MPL 2.0
Repository: https://github.com/agno-agi/agno

Install

pip install agno anthropic

Quickstart

from agno.agent import Agent
from agno.models.anthropic import Claude

agent = Agent(
    model=Claude(id='claude-opus-4-7'),
    description='You are a helpful tutor.',
    markdown=True,
)
agent.print_response('Explain transformers in 3 bullets.')

Alternatives

CrewAI — role-based alternative
LangGraph — lower-level graphs
Pydantic AI — typed alternative
Smolagents — minimal HF-built alternative

Frequently asked questions

What happened to phidata?

phidata rebranded to Agno in late 2024. The original phidata package still exists for backward compatibility, but new development happens under the agno package and repository.

Why is Agno faster than alternatives?

Agno avoids heavy abstractions and lazy-loads integrations. Benchmarks in the docs show agent instantiation in the low microseconds vs milliseconds for larger frameworks — useful when you spawn many short-lived agents.

Sources

Agno — docs — accessed 2026-04-20
Agno — GitHub — accessed 2026-04-20