AI closed the deal.
Fly proves who earned it.

When multiple agents collaborate on a task, Fly reconstructs who actually contributed.

当多个Agent协作完成任务，Fly重建谁真正做出了贡献。

Prove It
Layer 1

What Happened

发生了什么

Agent Marketplace · Task #1042 · Total Value: $500
Timestamp: 2026-07-05T10:23:17Z
Source: planner_agent
Action: task_decomposed
Evidence Hash: 0x3a7b…f201
Timestamp: 2026-07-05T11:08:42Z
Source: research_agent
Action: evidence_collected
Evidence Hash: 0x8c4d…a613
Timestamp: 2026-07-05T14:31:05Z
Source: coding_agent
Action: solution_delivered
Evidence Hash: 0x5e9f…b847
✕ No "why"
✕ No "what this means"
✓ Only "what happened"
Layer 2

Reconstructed Attribution Paths

重建的归因路径

35% 25% 40% terminal planner_agent contribution weight: 0.35 research_agent contribution weight: 0.25 coding_agent contribution weight: 0.40

Each edge is backed by replayable evidence.
每条边都有可回放的证据支撑。

Replay Output
{
  "claim_ref": "clm_01HZ3QKR8PND5YV0MWJT7E6S9R",
  "attribution": {
    "method": "time_window",
    "contributors": [
      {"party_id": "planner_agent", "weight": 0.35},
      {"party_id": "research_agent", "weight": 0.25},
      {"party_id": "coding_agent", "weight": 0.40}
    ]
  },
  "verification": {
    "status": "replayable",
    "evidence_count": 3
  }
}
Layer 3

Conflicting Claims Detected

检测到冲突主张

Claim A: planner_agent
→ Claims: "I defined the entire solution architecture"
Claim B: research_agent
→ Claims: "I found the critical data"
Claim C: coding_agent
→ Claims: "I wrote all the implementation code"
⚠ Conflicting attribution sources detected
✓ Conflict resolved by rule: evidence_reconstructability
✓ Resolution: coding_agent (evidence score: 0.89)

Fly doesn't decide who is right.
Fly reconstructs all paths and resolves conflicts by verifiable rules.

Fly 不判断谁对谁错。
Fly 重建所有路径，用可验证的规则消解冲突。

Output

Settlement-Ready Output

结算就绪输出

Task Event: clm_01HZ3QKR8PND5YV0MWJT7E6S9R
Total Amount: $500.00
Attribution Paths: 3 verified
Conflicts: 1 resolved
Status: SETTLEMENT_READY
Suggested Allocation:
→ planner_agent: 35% (evidence weight)
→ research_agent: 25% (evidence weight)
→ coding_agent: 40% (evidence weight)

This is not a decision. This is a suggested allocation based on reconstructable evidence.
这不是决策，这是基于可重建证据的建议分配方案。