The simplest open-source library for LLM agents. Granular control, zero hidden logic. Composable building blocks for intelligent systems.
from instantneo import InstantNeo
neo = InstantNeo(
provider="openai",
model="gpt-4o",
api_key="...",
role_setup="Eres Neo, el elegido, de la Matrix"
)
neo.run("¿Qué es la Matrix?")Inspired by Minsky's Society of Mind: specialized agents you compose to build intelligent systems. Each agent is transparent, controllable, and minimal.
Own your execution flow. WAIT_RESPONSE, EXECUTION_ONLY, or GET_ARGS—you decide every step. No hidden orchestration.
5 lines to start. One decorator for skills. Switch providers with one parameter. Built for speed, not ceremony.
See what happens, when it happens. Every action is explicit and traceable. No framework magic.
Install and teach Neo kung fu in 90 seconds.
pip install instantneo[all]
# or target specific providers
# pip install instantneo[openai]
# pip install instantneo[anthropic]
from instantneo import InstantNeo
from instantneo.skills import skill
# 1. Define una skill
@skill(description="""Ejecuta movimientos de Kung Fu para defenderte de ataques:
Puño recto, Garra de tigre, Garra del dragón, Patada tornado""")
def kung_fu(movimiento: str, intensidad: int = 5):
return f"¡Ejecutando {move} con intensidad {intensity}!"
# 2. Create a Neo con skills
neo = InstantNeo(
provider="anthropic",
model="claude-3-7-sonnet-20250219",
api_key="...",
role_setup="Eres Neo, el elegido, de la Matrix",
skills=[kung_fu]
)
# 3. Run
neo.run("Neo, te atancan tres agentes por el frente")Switch between OpenAI, Anthropic, Groq, and others with a consistent API. The library hides provider quirks.
Turn functions into skills via a decorator. Add metadata, validation, and compose dynamically.
Choose execution modes: WAIT_RESPONSE for full cycles, EXECUTION_ONLY for tools, or GET_ARGS for planning.
Early contributors shape the core. Fast merges, direct impact.