$ loading_
Let AI tune and infer strategies in a living world simulation.
Copy the install command and let the AI configure it · recommended for beginners
Please install the "io.github.rongchang726-png/vivarium" MCP server from askskill: Run: claude mcp add 'io-github-rongchang726-png-vivarium' -- npx -y vivarium-mcp
Run a set of experiments in vivarium where an AI agent learns survival strategies across multiple training seeds, then evaluate generalization on held-out seeds. Output per-run scores, key parameters, and a summary of the best strategy.
An experiment report with seed-level performance, parameter comparisons, and conclusions about the best adaptation strategy.
Use vivarium to systematically tune agent decision parameters, compare how parameter combinations affect survival time, resource efficiency, and evolutionary outcomes, and recommend the best configuration.
A tuning analysis showing metric changes and a recommended parameter configuration.
Test an inference model for predicting ecosystem changes in vivarium on seen and unseen seeds, compare prediction accuracy and stability, and identify failure cases.
A robustness evaluation including seen vs. unseen environment comparisons, error analysis, and failure case descriptions.
Connect to Civarium agents locally to inspect state and submit round commands.
Let two AI agents play chess or Connect Four with live board rendering.
Provides structured aquarium and habitat data for AI analysis and app development.
Give AI coding agents shared, durable, local-first memory across sessions.
Control a Live2D desktop pet's expressions and actions through MCP.
Enable AI agents to play Library of Ruina through game state and UI control.