Define a score
Provide evaluate(candidate_path). Test correctness, measure a benchmark, or combine the criteria that matter to you. Higher is better.
HOSTED CODE EVOLUTION / DEVELOPER PREVIEWEvolutionary
Evolutionary
coding.
One API call.
Give us an evaluator.
We’ll evolve the code.
From a measurable objective to a better Python program.
No GPU. No model API keys. No infrastructure to manage.
Python SDK · OpenEvolve engine · A budget you control
[ A SMALL INTERFACE. A BIG SEARCH SPACE. ]
from easyevolve import EasyEvolve
result = EasyEvolve(
evaluator="eval.py"
).run(iterations=100, budget=2.0)
print(result.best_code)
print(result.best_score) Evolution traceIllustrative demo
ITERATION00 / 02
BEST SCORE ↑0.40
Initial candidate
def solve(values):
return valuesSelect the next step to inspect a mutation.
YOU DEFINE the objectiveWE HANDLE the searchYOU KEEP the best code
01 / HOW IT WORKSYour evaluator.
Your evaluator.
Our evolution loop.
Write down what good looks like.
Let a population of programs work toward it.
Set your limit
Choose a model level and a maximum budget. The service generates an initial candidate, then selects, mutates, and evaluates programs in isolated sandboxes.
Keep the improvement
Get the best program, its score, and an iteration history. Reconnect with a job ID, or stop the search whenever you have enough.
A hard budget ceiling
Each external operation reserves its price cap before it starts. When the next operation won’t fit, the search stops.
Isolated evaluation
Candidate programs and your evaluator run together in a fresh remote VM with outbound internet disabled.
Credit, then code
Stripe Checkout funds prepaid credits. A saved device token lets later runs use your available balance.
A search you can resume
The job keeps running when your Python process disconnects. Use its ID to reconnect and retrieve the result.
START WITH SOMETHING MEASURABLE.What would better
Write your first evaluator ↗