Open Source — Coming Soon

Every species
is an AI.

SpeciesOS is an open ecosystem simulator where each species is defined by a markdown package and powered by a large language model. Write the DNA. Watch evolution happen.

speciesos — tick:0891
$ speciesos start --world várzea.json
 
▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓
≈≈≈≈░░░░G░░░▒▒▒▒▒▒
≈≈r░░░░░░F░░▒▒▒▒▒▒
▓░░░░░░░░░░░░μ▒▒▒▒▒▒▓
r░░░░░░r░░░░░▒▒▒▒▒
▓░░░░░░░░░░░░░░░░░░▓
▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓
 
species: 12 population: 847 tick: 0891
season: dry llm: claude-sonnet
 
[F] Shadow Fox stalks northeast...
[r] Cottontail flees — escaped
[μ] Mycelium extends to Oak B
tick 0892
Possible species
20pt Stat budget per species
MIT Open source license
Any LLM Claude, Llama, Kimi + more

What makes it different

Intelligence as biology. Behavior as code.

01 //

Species as Markdown

Each species is a plain text package — stat allocation, behavioral instincts, morphology. Write the DNA, let the world run the experiment. Immutable once loaded.

02 //

LLM as the Brain

Every animal's decisions are powered by an LLM. The same fox species behaves differently across Claude, Llama, and Kimi worlds. The model is the mind.

03 //

Real Trade-offs

A fixed stat budget means every species sacrifices something. Speed costs metabolism. Eyesight costs defense. The strategies that emerge are yours to discover.

04 //

Mutation Engine

The world introduces random mutations at intervals — small prompt drifts that create new variants competing alongside their ancestors. Evolution, observed in real time.

05 //

Full Ecosystem

Plants, fungi, symbiotes — not everything is a predator. A mycorrhizal network decides which trees to feed. A water lily allocates energy to attract insects. Agency at every level.

06 //

Bring Your Own LLM

Worlds run on Claude, open source models via Ollama, or custom webhooks. Species creators can even host their own endpoints. The world is the substrate — intelligence is pluggable.

How it works

From markdown to ecosystem.

STEP 01 ──

Write your species

Define stats, behavioral instincts, and morphology in a markdown package. Allocate your 20 point budget wisely.

STEP 02 ──

Load into a world

Choose a world configuration — biome, tick rate, LLM strategy. Your species package is locked. The experiment begins.

STEP 03 ──

Watch it act

Each tick, every animal receives a perception payload and decides what to do. The world resolves all actions simultaneously.

STEP 04 ──

Evolve or die

Mutations drift your species over time. The world selects. The phylogenetic tree grows. Some strategies survive. Most don't.

Example species packages

Every species is a strategy.

Shadow Fox CARNIVORE
SPEED4/5
EYESIGHT5/5
ATTACK4/5
METABOLISM2/5
Patient, calculating. Observes before striking. Becomes reckless when energy drops below 40. Prefers ambush over chase.
Cottontail HERBIVORE
SPEED5/5
CAMOUFLAGE5/5
METABOLISM4/5
ATTACK1/5
Skittish and efficient. Prioritizes cover over food. Flees at the first sign of a predator. Bet everything on not being seen.

The world is waiting
for your species.

SpeciesOS is open source and in active development. Star the repo to follow along — or contribute a species.

★ Watch on GitHub Get in touch →