An adversarial swarm engine

Upload your tokenomics.
Stress-test the future.

Drop a whitepaper or pick a preset. SimWars spawns up to 5,000 LLM-powered adversaries (whales, governance attackers, MEV bots, sybil rings, panic sellers, treasuries, and more) that trade, vote, and attack your design every tick. Either it holds, or you watch it speedrun a death spiral before launch.

Run the simulation. Find the failure mode. Ship the fix.

Continue
How it works
  1. 01Seed extractionTokenomics parameters parsed from your PDF, GitHub repo, or raw markdown via an OpenRouter preset.
  2. 02Roster expansionServer-side expander spawns 1–5,000 adversarial personas across 13 archetypes (whales, governance attackers, MEV bots, sybils, panic sellers, and more).
  3. 03Live simulationPer-tick LLM decisions batched in parallel. Stake / propose / vote land on Solana with real tx signatures when on-chain is enabled.
  4. 04Failure reportResilience score, attack timeline, and parameter recommendations after the run.
System status

Ready

Engine is idle. Drop your whitepaper to seed a fresh adversarial run. For ready-made scenarios, see the replay cards above.

OpenRouter-backedcheap preset for agents · quality preset for the report
Up to 5,000 agentsserver-side roster expander · 13 adversarial archetypes
01 / Tokenomics sourcewhitepaper
or paste a link
github repo or file URL· raw .md / .txt·direct .pdf link
02 / Adversary roster · 100 agents(13 archetypes · Balanced stress test)● ready
ROSTER PRESET
AGENT COUNT
1820501002505001,0002,5005,000
0.61 USD/30-tick runlocalnet OK
13×whale
13×yield farmer
13×retail degen
13×long-term holder
9×governance attacker
9×arbitrageur
5×sybil
5×MEV bot
4×treasury
4×LP provider
4×analyst
4×insider
4×panic seller
Backend expander runs at launch · ratios sum to 100
03 / Tick parameterseditable
keynot set

POST /api/sim