Multi-Agent Mindustry Opens Agent Benchmarks
Olam Labs open-sources Multi-Agent Mindustry, a benchmark suite for persistent-agent PvP series and cooperative endless survival scenarios built on Mindustry and Harbor Multiagent. Developers can configure teams, models, seat counts, and replayable runs locally. ([YC Labs](https://labs.ycombinator.com/projects/games/), [GitHub](https://github.com/olamlabs/multi-agent-mindustry))
This is a practical testbed for long-horizon coordination, but its real value will depend on reproducible runs across models rather than one-off game outcomes.
- –PvP supports 1–100 consecutive games with persistent sessions, enabling adaptation and strategy analysis across a series.
- –Endless Survival supports up to 20 cooperating agents defending against enemy waves.
- –Equal game APIs and configurable harnesses make it useful for studying specialization, communication, and emergent team structures.
- –Built-in recordings, replay reconstruction, regression tests, and integration checks make results easier to inspect and reproduce.
- –The GPL-3.0 license and Docker-based setup lower experimentation barriers, though compute requirements remain significant. ([GitHub](https://github.com/olamlabs/multi-agent-mindustry))
DISCOVERED
1h ago
2026-11-01
PUBLISHED
1h ago
2026-11-01
RELEVANCE
AUTHOR
AI Revolution
