{
  "title": "Prompt cache telemetry",
  "description": "Report cache reads, writes, hits, misses, saved latency, or cached-token billing for a run.",
  "slug": "prompt-cache-telemetry",
  "locale": "en",
  "seoTitle": "Prompt cache telemetry compatibility — Can My Agent Use",
  "socialTitle": "Prompt cache telemetry",
  "socialDescription": "Compare visibility into prompt-cache hits, misses, token counts, latency, and cost.",
  "llmSummary": "Prompt-cache telemetry exposes whether reuse occurred and, ideally, cached read/write tokens, cost, latency, and cache scope; documented caching without visibility is a separate claim.",
  "audience": "Engineers verifying that context caching is effective and economical.",
  "contentKind": "feature",
  "status": "published",
  "tags": [
    "context",
    "prompt-caching",
    "telemetry",
    "cost"
  ],
  "updated": "2026-08-28T00:00:00.000Z",
  "published": "2026-08-28T00:00:00.000Z",
  "category": "models-context",
  "summary": "Inspect cache hits, misses, token counts, latency, or billing effects.",
  "specLabel": "Common product term",
  "aliases": [
    "cache hit metrics",
    "cached tokens",
    "cache observability"
  ],
  "capabilityKind": "atomic",
  "parent": "models-and-context",
  "related": [
    "prompt-cache-reuse",
    "prompt-cache-controls",
    "usage-metering"
  ],
  "relations": [],
  "highlight": false,
  "notes": [
    {
      "id": 1,
      "text": "Evidence checked 2026-08-28: Claude Code exposes cache creation and cache read token counts through status-line data and OpenTelemetry, including per-user and per-session organization metrics."
    },
    {
      "id": 2,
      "text": "Evidence checked 2026-08-28: Gemini CLI documents cached-token savings in /stats when token caching is available."
    }
  ],
  "issues": [],
  "resources": [
    {
      "title": "Methodology",
      "href": "/methodology",
      "kind": "note"
    },
    {
      "id": "anthropic-code-prompt-caching",
      "title": "Anthropic — How Claude Code uses prompt caching",
      "href": "https://code.claude.com/docs/en/prompt-caching",
      "kind": "docs",
      "publisher": "Anthropic",
      "evidenceType": "documented",
      "reviewedAt": "2026-08-28"
    },
    {
      "id": "anthropic-code-monitoring",
      "title": "Anthropic — Monitor Claude Code usage",
      "href": "https://code.claude.com/docs/en/monitoring-usage",
      "kind": "docs",
      "publisher": "Anthropic",
      "evidenceType": "documented",
      "reviewedAt": "2026-08-28"
    },
    {
      "id": "google-gemini-cli-token-caching",
      "title": "Google — Gemini CLI token caching",
      "href": "https://geminicli.com/docs/cli/token-caching/",
      "kind": "docs",
      "publisher": "Google",
      "evidenceType": "documented",
      "reviewedAt": "2026-08-28"
    }
  ],
  "support": [
    {
      "harness": "claude-cli",
      "versions": [
        {
          "track": "current",
          "status": "yes",
          "noteIds": [
            1
          ],
          "target": {
            "kind": "dated-documentation",
            "revision": "current Claude Code documentation",
            "observedAt": "2026-08-28"
          },
          "environmentProfile": "local-default",
          "qualifiers": [
            {
              "type": "runtime",
              "value": "current_usage exposes cache_creation_input_tokens and cache_read_input_tokens; OpenTelemetry reports cache read and creation tokens per request, user, and session"
            }
          ],
          "evidence": [
            {
              "resourceId": "anthropic-code-prompt-caching",
              "type": "documented",
              "observedAt": "2026-08-28"
            },
            {
              "resourceId": "anthropic-code-monitoring",
              "type": "documented",
              "observedAt": "2026-08-28"
            }
          ]
        }
      ]
    },
    {
      "harness": "gemini-cli",
      "versions": [
        {
          "track": "current",
          "status": "partial",
          "noteIds": [
            2
          ],
          "target": {
            "kind": "dated-documentation",
            "revision": "current Gemini CLI documentation",
            "observedAt": "2026-08-28"
          },
          "environmentProfile": "local-default",
          "qualifiers": [
            {
              "type": "runtime",
              "value": "/stats displays cached-token savings when caching is available; the reviewed page does not establish hit/miss events or retained historical telemetry"
            }
          ],
          "evidence": [
            {
              "resourceId": "google-gemini-cli-token-caching",
              "type": "documented",
              "observedAt": "2026-08-28"
            }
          ]
        }
      ]
    }
  ],
  "html": "/features/prompt-cache-telemetry",
  "markdown": "/features/prompt-cache-telemetry.md",
  "json": "/api/v1/features/prompt-cache-telemetry.json",
  "body": "This row asks whether an operator can verify prompt-cache behavior for the exact harness. A billing dashboard total, per-request cached-token count, cache-hit event, and detailed read/write breakdown provide different levels of evidence and should be qualified.\n\nUseful records identify where telemetry appears, how long it is retained, whether sub-agent and tool-generated requests are included, and how cached reads and writes map to price. A generic statement that caching is automatic does not prove that the harness exposes telemetry."
}