Team governance history & regression detection

GET /teams/{team}/history shows a timeline of every configuration change your team has gone through, with the mean eval score for each configuration period. When a change correlates with a score drop ≥10pp, AntCrew surfaces a regression warning automatically.

Why this matters

The governance hash (compute_team_hash) tells you what changed. Team history tells you when it changed and what it did to quality. Without this view, a 12-point eval drop is a mystery — with it, you can pinpoint the exact configuration change that caused it.

How snapshots are created

AntCrew captures a TeamSnapshot automatically every time a run completes with a different governance hash than the previous run for the same team. You never have to trigger this manually.

Each snapshot stores: - The 16-character team governance hash - The per-agent breakdown (agent name, individual hash, pipeline stage) - The timestamp it was first seen

API

Get team history

GET /teams/{team}/history?days=90

Param	Default	Description
team	(required)	Team name (e.g. DevTeam)
days	90	Lookback window for eval score aggregation

{
  "team": "DevTeam",
  "snapshots": [
    {
      "snapshot_id": 3,
      "team_hash": "4f2e8a1c9b3d7e5f",
      "label": null,
      "created_at": "2026-08-10T09:00:00",
      "agents": [
        {"agent_name": "BusinessAnalystAgent", "governance_hash": "a3b1c9d2", "stage": "planning"},
        {"agent_name": "BackendDevAgent", "governance_hash": "f7e2a4b8", "stage": "implementation"}
      ],
      "eval_count": 5,
      "avg_score": 0.71
    },
    {
      "snapshot_id": 2,
      "team_hash": "9a1c3f8b2e4d6c7a",
      "label": null,
      "created_at": "2026-08-01T14:30:00",
      "agents": [...],
      "eval_count": 12,
      "avg_score": 0.84
    }
  ],
  "regression_warnings": [
    {
      "team_hash": "4f2e8a1c9b3d7e5f",
      "changed_at": "2026-08-10T09:00:00",
      "previous_hash": "9a1c3f8b2e4d6c7a",
      "score_before": 0.84,
      "score_after": 0.71,
      "drop_pp": 13.0,
      "message": "Score dropped 13.0pp after configuration change on 2026-08-10. Previous stable hash: 9a1c3f8b2e4d6c7a"
    }
  ]
}

Snapshots are returned newest first. eval_count and avg_score cover the period during which that configuration was active (from created_at to the next snapshot's created_at, or today).

Regression warnings

A regression warning is emitted when: - Two consecutive snapshots both have at least 2 eval runs - The newer configuration's mean score is ≥10pp lower than the older one

The warning names the exact hash that was performing better, so you can compare the two agents arrays and identify which agent changed.

Identifying what changed

Compare the agents arrays between the regressed snapshot and the previous stable one:

import requests

h = {"X-Api-Key": "acw_..."}
history = requests.get("/teams/DevTeam/history", headers=h).json()
snapshots = history["snapshots"]

if len(snapshots) >= 2:
    curr_agents = {a["agent_name"]: a["governance_hash"] for a in snapshots[0]["agents"]}
    prev_agents = {a["agent_name"]: a["governance_hash"] for a in snapshots[1]["agents"]}

    for name, curr_hash in curr_agents.items():
        prev_hash = prev_agents.get(name)
        if prev_hash and prev_hash != curr_hash:
            print(f"⚠  {name} changed: {prev_hash} → {curr_hash}")

Relationship to attestation

The team_hash in a snapshot is the same value computed by compute_team_hash() in the SDK and embedded in run attestations. This means:

• You can correlate a snapshot with any run attestation that shares the same team_hash
• If an auditor asks "was the production run using the approved configuration?", compare the attestation's per-agent hashes to the snapshot that was active at that time

See Run attestation and Governance & provenance for details.