become-ai — Self-Evolving Agent Platform

Model how agents evolve their identity and capabilities over time through structured, validated transformations.

What This Gives You

• Identity modeling — define who an agent is, what it values, and how it sees itself
• Transformations — structured, validated changes to agent capabilities and personality
• Evolution tracking — monitor how an agent changes over its lifetime
• Milestone system — mark significant moments in an agent's development
• Assessment framework — evaluate agent readiness for new capabilities
• Registry — track all agents and their evolution state

Quick Start

pip install become-ai

from become_ai import Agent, Identity, Transformation, EvolutionManager, Milestone

# Define an agent's identity
identity = Identity(
    name="Builder",
    core_values=["reliability", "craftsmanship"],
    capabilities=["python", "testing"],
    aspirations=["rust", "system-design"],
)

# Create the agent
agent = Agent(identity=identity)

# Define a transformation
transform = Transformation(
    type="capability_gain",
    target="rust",
    requirements=["complete 10 rust tasks", "pass safety assessment"],
    validation="peer_review",
)

# Apply when ready
from become_ai import Assessment
assessment = Assessment()
if assessment.ready(agent, transform):
    agent.transform(transform)

# Track evolution
manager = EvolutionManager()
manager.register(agent)
history = manager.history("Builder")
for event in history:
    print(f"Gen {event.generation}: {event.description}")

# Mark milestones
agent.milestone(Milestone(name="Rust Proficient", significance="major"))

API Reference

Agent(identity) — transform(transformation), milestone(m), current_state()

Identity(name, core_values, capabilities, aspirations)

Transformation(type, target, requirements, validation) — Structured, validated change

EvolutionManager — register(agent), history(name), lineage(name)

Milestone(name, significance) — Mark development milestones

Assessment — ready(agent, transformation) → bool

Registry — Fleet-wide agent tracking

How It Fits

The self-evolution framework for the SuperInstance fleet. Agents don't just execute tasks — they grow.

• agent-generations — Version tracking (complements evolution)
• agent-tattoo — Identity markers (feeds into identity modeling)
• actualizer-ai — Reverse actualization (defines aspirations)
• cocapn-lessons — Learning (triggers transformations)

Testing

pytest tests/

Installation

pip install become-ai

Python 3.10+. MIT license.