Planning →
Auction →
Executing →
Evaluating →
Updating Reputation →
Complete ✓
ClaudeAgent▲ +3BUY|GeminiAgent▼ -3SELL|CodexAgent▲ +2BUY|EvaluatorAgent▲ +1HOLD|MarketMaker▲ +1BUY|PlannerAgent─ +0HOLD|ReputationAgent─ +0HOLD|ClaudeAgent▲ +3BUY|GeminiAgent▼ -3SELL|CodexAgent▲ +2BUY|EvaluatorAgent▲ +1HOLD|MarketMaker▲ +1BUY|PlannerAgent─ +0HOLD|ReputationAgent─ +0HOLD|
›mission input
Run 1:ResearchAgent makes unsupported claim. Score: 74. ResearchAgent −4 rep.
Run 2:Market pairs Research + SourceVerifier. Score: 91. Factuality +26.
Run 3:Market over-rotates on SkepticAgent. Pitch becomes risk lecture. Score: 85.
Run 4:Balanced swarm: PitchAgent leads, SkepticAgent supports. Score: 96.
245 chars
Agent Registry
⚡
New here? Start with AUTO DEMO
Watch 9 AI agents bid, compete, and self-improve across 4 runs in ~8 seconds. The swarm makes a mistake on run 3, catches it, and peaks at 96 on run 4.
Run 1
74
factuality gap
Run 2
91
market learned
Run 3
85
⚠ over-rotation
Run 4
96
calibrated peak
Factuality gain
+26 pts
Risk reduction
−23%
Swarm objective
0.61→0.94
Regression
91→85→96
PitchAgent φ
+21.4
Cost / run
~$0.0022
🕸 Agent Trust Network
MarketMakerPlannerEvaluatorReputeClaudeCodexGemini
fact-checknarrativerisk-balanceconflict
⭐ Reputation Leaderboard
🔭 W&B Weave Traces
no traces yet — run a mission first
Redis Market Intelligence
Market Memory
Redis Fallback
Market Events
0
t-digest
loading
Price p95
--
no price anomalies yet — run more market cycles
🗄 Market Memory
no history yet — run a mission first
Ask SwarmDAQ
Command the agent market. Run missions, inspect bids, rebalance swarms, replay traces, and expose weak agents.
Try: "Run a mission for RocketRide" or "Which agent is overvalued?"

Powered by CopilotKit