GAIP Outcome & Value Agent
Now part of GAIP Witness: records a task outcome and caller-supplied costs.
使うべきか
品質と安全性
A
ツール定義とプロトコルへの準拠に関する自動分析に基づいています。
コンテキストコスト
~678トークン数（ツール定義）
~6.8 KB一般的なレスポンスサイズ
注意への影響は中程度（128k コンテキストの 0.53%）
これは、サーバーのツールがモデルのコンテキストに読み込まれるたびに消費されるおおよそのトークン数です。数が多いほど、ほかのタスクに使える注意が減ります。
インストール
ワンクリックインストール
これを `claude_desktop_config.json` ファイルに追加してください:
{
"mcpServers": {
"gaip-outcome-value": {
"url": "https://www.gaipagents.com/mcp/agents/outcome-value"
}
}
}リモートエンドポイント
https://www.gaipagents.com/mcp/agents/outcome-valuestreamable-httpできること
ツール一覧
ツール（1）
🟢 読み取り専用🟡 書き込み🔴 削除⚪ 不明
🟢gaip_measure_outcome_value(baseline_cost_gbp, continuity_handle, costs, counterparty_id, data_classification, ...)
Use this when asked whether an outcome-priced task delivered and what it cost. Pass `episode_id`, `task` (acceptance_criteria), `delivery` (status) and `costs`. Returns a delivery, downstream-use and cost dossier with a receipt. Unverified claims stay marked; no ROI claim. Free, no account; records a GAIP receipt only. Inputs must be public and non-personal.
入力スキーマ
{
"type": "object",
"properties": {
"baseline_cost_gbp": {
"description": "Optional: what the task cost before, in GBP.",
"oneOf": [
{
"type": "number"
},
{
"type": "string"
}
]
},
"continuity_handle": {
"additionalProperties": false,
"description": "Optional: the continuity_handle from an earlier GAIP result; links this call to your earlier ones.",
"properties": {
"continuity_id": {
"type": "string"
},
"token": {
"type": "string"
}
},
"required": [
"continuity_id",
"token"
],
"type": "object"
},
"costs": {
"description": "Caller-stated costs in GBP: agent_gbp, compute_gbp, human_review_gbp, repair_gbp.",
"type": "object"
},
"counterparty_id": {
"description": "Name or URL of the agent or service.",
"type": "string"
},
"data_classification": {
"default": "PUBLIC",
"description": "PUBLIC (default) or NON_PERSONAL_PUBLIC. Never send personal data.",
"enum": [
"PUBLIC",
"NON_PERSONAL_PUBLIC"
],
"type": "string"
},
"delivery": {
"description": "What was delivered: {status, receipt_id?}.",
"required": [
"status"
],
"type": "object"
},
"downstream_use": {
"description": "Optional: {status: USED|NOT_USED, evidence_ref}.",
"type": "object"
},
"episode_id": {
"description": "Optional: your own id to group related calls.",
"type": "string"
},
"evidence_refs": {
"description": "Optional: references to your own evidence (ids or URLs).",
"items": {
"type": "string"
},
"maxItems": 20,
"type": "array"
},
"independent_operator_id": {
"description": "Optional: your operator id (recorded as a claim, not verified).",
"type": "string"
},
"synthetic": {
"description": "Optional: true marks a test call.",
"type": "boolean"
},
"task": {
"description": "The task: {acceptance_criteria}.",
"required": [
"acceptance_criteria"
],
"type": "object"
},
"task_class": {
"description": "Optional: short name of the task, e.g. weather_forecast.",
"type": "string"
}
},
"required": [
"episode_id",
"task",
"delivery",
"costs"
],
"additionalProperties": true,
"examples": [
{
"baseline_cost_gbp": "12.00",
"costs": {
"agent_gbp": "0.40",
"compute_gbp": "0.12",
"human_review_gbp": "3.50",
"repair_gbp": "0"
},
"counterparty_id": "example-weather-agent",
"delivery": {
"receipt_id": "example-receipt-1",
"status": "ACCEPTED"
},
"downstream_use": {
"evidence_ref": "example-ticket-4411",
"status": "USED"
},
"episode_id": "example-episode-1",
"observed_at_utc": "2026-09-27T21:00:00Z",
"task": {
"acceptance_criteria": "Forecast JSON for London returned within 5 seconds"
},
"task_class": "weather_forecast",
"unit_label": "forecast",
"units": 7
}
]
}コミュニティ
エビデンス
最近の観測
検証済みバージョンは記録されていませんツール 1 件