Add Claude Enterprise Analytics integration (unofficial)

.codecov.yml

@@ -23,6 +23,10 @@ coverage:
         target: 75
         flags:
         - bind9
+      Claude_Enterprise_Analytics:
+        target: 75
+        flags:
+        - claude_enterprise_analytics
       CloudNatix:
         target: 75
         flags:
@@ -374,6 +378,11 @@ flags:
     paths:
     - cfssl/datadog_checks/cfssl
     - cfssl/tests
+  claude_enterprise_analytics:
+    carryforward: true
+    paths:
+    - claude_enterprise_analytics/datadog_checks/claude_enterprise_analytics
+    - claude_enterprise_analytics/tests
   cloudnatix:
     carryforward: true
     paths:

.github/CODEOWNERS

@@ -1183,6 +1183,8 @@ code-coverage.datadog.yml                                  @DataDog/agent-integr
 /dagster/              @DataDog/ecosystems-review
 /aerospike_enterprise/ @support-aerospike @DataDog/ecosystems-review
 /scamalytics/          @ScamalyticsDev @DataDog/ecosystems-review
+/claude_enterprise_analytics/                @matiasozdy @DataDog/ecosystems-review
+/claude_enterprise_analytics/*metadata.csv   @matiasozdy @DataDog/documentation @DataDog/ecosystems-review
 
 # LEAVE THE FOLLOWING LOG OWNERSHIP LAST IN THE FILE
 # Make sure logs team is the full owner for all logs related files

.github/workflows/test-all.yml

@@ -158,6 +158,25 @@ jobs:
       test-py3: ${{ inputs.test-py3 }}
       setup-env-vars: "${{ inputs.setup-env-vars }}"
     secrets: inherit
+  jf0663d2:
+    uses: DataDog/integrations-core/.github/workflows/test-target.yml@574d63ba88365ffbab915280ceddbaa333c63d6a
+    with:
+      job-name: Claude Enterprise Analytics
+      target: claude_enterprise_analytics
+      platform: linux
+      runner: '["ubuntu-22.04"]'
+      repo: "${{ inputs.repo }}"
+      context: ${{ inputs.context }}
+      python-version: "${{ inputs.python-version }}"
+      latest: ${{ inputs.latest }}
+      agent-image: "${{ inputs.agent-image }}"
+      agent-image-py2: "${{ inputs.agent-image-py2 }}"
+      agent-image-windows: "${{ inputs.agent-image-windows }}"
+      agent-image-windows-py2: "${{ inputs.agent-image-windows-py2 }}"
+      test-py2: ${{ inputs.test-py2 }}
+      test-py3: ${{ inputs.test-py3 }}
+      setup-env-vars: "${{ inputs.setup-env-vars }}"
+    secrets: inherit
   j6a8ad70:
     uses: DataDog/integrations-core/.github/workflows/test-target.yml@574d63ba88365ffbab915280ceddbaa333c63d6a
     with:

claude_enterprise_analytics/CHANGELOG.md

@@ -0,0 +1,4 @@
+# CHANGELOG - Claude Enterprise Analytics
+
+<!-- towncrier release notes start -->
+

claude_enterprise_analytics/README.md

@@ -0,0 +1,94 @@
+# Agent Check: Claude Enterprise Analytics
+
+> **Unofficial integration.** This check is community-maintained and **not affiliated with, endorsed by, or supported by Anthropic, PBC**. "Anthropic" and "Claude" are trademarks of Anthropic, PBC. For official support of the underlying API, contact Anthropic.
+
+## Overview
+
+This integration pulls daily usage, cost, and seat-utilization data from the [Anthropic Claude Enterprise Analytics API][1] and submits it to Datadog as metrics. It is useful for Anthropic enterprise customers who want Claude usage observability inside their existing Datadog dashboards: cost attribution per user/model/product, seat adoption tracking, Claude Code activity (commits, PRs, lines, tool acceptance rate), token mix and cache effectiveness, and web-search usage.
+
+The check polls six API endpoints once per collection interval for a single past day (default: today minus 3, because Anthropic's analytics have a 3-day aggregation lag):
+
+- `/summaries` -- org-wide DAU/WAU/MAU, assigned seats, adoption rates
+- `/users` -- per-user activity across chat, projects, artifacts, Claude Code, web search
+- `/usage_report` -- token consumption by model and product
+- `/cost_report` -- USD spend by model and product (list price; see "Known Limitations")
+- `/user_usage_report` -- per-user token totals
+- `/user_cost_report` -- per-user USD spend
+
+All metrics are emitted as gauges at agent wall-clock time and tagged with `report_date:YYYY-MM-DD` representing the activity day. Build dashboards that filter or group by `report_date` rather than relying on Datadog's wall-clock x-axis.
+
+## Setup
+
+### Installation
+
+1. Install the [Datadog Agent][2] on your host.
+2. Install this integration from the `integrations-extras` repo:
+   ```
+   datadog-agent integration install -t datadog-claude-enterprise-analytics==<version>
+   ```
+
+### Configuration
+
+1. Generate an API key at [claude.ai/analytics/api-keys][1] (requires the Primary Owner role on your Anthropic org). The key must carry the `read:analytics` scope.
+2. Edit `claude_enterprise_analytics.d/conf.yaml` in your Agent's `conf.d/` directory:
+   ```yaml
+   instances:
+     - anthropic_api_key: <YOUR_KEY>
+       org_id: my-org             # free-form tag value, surfaced on every metric
+       lag_days: 3                 # how many days behind today to poll
+       min_collection_interval: 3600  # once per hour is plenty; Anthropic refreshes daily
+   ```
+   See [`conf.yaml.example`][4] for all options.
+3. [Restart the Agent][5].
+
+### Validation
+
+Run [`datadog-agent status`][6] and look for `claude_enterprise_analytics` under the Checks section. Then in the Datadog UI, open Metrics Explorer and search for `claude_enterprise_analytics.*` -- you should see ~45 metrics arrive within one collection interval.
+
+## Data Collected
+
+### Metrics
+
+45 gauges under the `claude_enterprise_analytics.` namespace. Highlights:
+
+- `claude_enterprise_analytics.org.dau` / `.wau` / `.mau` / `.seats_assigned` / `.adoption_rate.{daily,weekly,monthly}`
+- `claude_enterprise_analytics.cost.amount_usd` / `.list_amount_usd` (tagged by `model`, `product`)
+- `claude_enterprise_analytics.user.cost.amount_usd` (tagged by `user_email`)
+- `claude_enterprise_analytics.tokens.{uncached_input,output,cache_read,cache_write_1h,cache_write_5m}`
+- `claude_enterprise_analytics.user.claude_code.{sessions,commits,prs,lines_added,lines_removed,tool_actions}`
+- `claude_enterprise_analytics.web_search_requests`
+
+See [`metadata.csv`][7] for the full list with descriptions and tag keys.
+
+### Events
+
+This integration does not submit events.
+
+### Service Checks
+
+- `claude_enterprise_analytics.can_connect` -- `CRITICAL` if the Anthropic API could not be reached during the check run, otherwise `OK`. See [`service_checks.json`][8].
+
+## Known Limitations
+
+- **Cost is list-priced.** Anthropic's API returns `amount` equal to `list_amount` in every response. Enterprise contract discounts are not exposed through this API, so the dashboard's USD figures reflect retail compute consumption rather than your actual invoice. Treat cost data as a relative attribution signal (which model/user is most expensive) rather than as billing truth.
+- **3-day data lag.** Each report represents activity from `today - 3` (configurable via `lag_days`). Polling more frequently than once per day will resubmit the same daily totals.
+- **Datadog metric submission window.** Metrics are submitted at the agent's current wall-clock time, not the activity date -- Datadog rejects timestamps more than ~1h in the past. The activity date is preserved via the `report_date` tag; dashboards must group/filter by it.
+- **Cost is returned in USD cents** (smallest currency subunit) despite the `currency: USD` field. This check divides by 100 before submission.
+
+## Troubleshooting
+
+- **No data after a fresh install:** Anthropic's analytics endpoint can return empty bodies for very recent dates. Try increasing `lag_days` from 3 to 5.
+- **`401 Unauthorized`:** confirm the API key is generated by a Primary Owner and carries the `read:analytics` scope.
+- **`400 Bad Request`** on the first run: the underlying analytics endpoints expect strict-before date ranges; this check handles that automatically, but a mismatched system clock can produce the wrong date. Verify the agent host has accurate UTC time.
+
+Need help with the integration itself? Open an issue against the [integrations-extras repository][10]. For Anthropic API issues, contact your Anthropic account team.
+
+[1]: https://support.claude.com/en/articles/13703965-claude-enterprise-analytics-api-reference-guide
+[2]: https://app.datadoghq.com/account/settings/agent/latest
+[3]: https://docs.datadoghq.com/containers/kubernetes/integrations/
+[4]: https://github.com/DataDog/integrations-extras/blob/master/claude_enterprise_analytics/datadog_checks/claude_enterprise_analytics/data/conf.yaml.example
+[5]: https://docs.datadoghq.com/agent/configuration/agent-commands/#start-stop-and-restart-the-agent
+[6]: https://docs.datadoghq.com/agent/configuration/agent-commands/#agent-status-and-information
+[7]: https://github.com/DataDog/integrations-extras/blob/master/claude_enterprise_analytics/metadata.csv
+[8]: https://github.com/DataDog/integrations-extras/blob/master/claude_enterprise_analytics/assets/service_checks.json
+[10]: https://github.com/DataDog/integrations-extras/issues

claude_enterprise_analytics/assets/configuration/spec.yaml

@@ -0,0 +1,38 @@
+name: Claude Enterprise Analytics
+files:
+- name: claude_enterprise_analytics.yaml
+  options:
+  - template: init_config
+    options:
+    - template: init_config/default
+  - template: instances
+    options:
+    - name: anthropic_api_key
+      required: true
+      secret: true
+      description: |-
+        API key for the Anthropic Claude Enterprise Analytics API. Generate one at
+        https://claude.ai/analytics/api-keys (requires Primary Owner role). Must
+        carry the `read:analytics` scope.
+      value:
+        type: string
+        example: <ANTHROPIC_ANALYTICS_API_KEY>
+    - name: org_id
+      required: false
+      description: |-
+        Free-form organization identifier emitted as the `org_id` tag on every
+        metric. Useful when more than one Anthropic org reports into the same
+        Datadog account.
+      value:
+        type: string
+        example: my-org
+    - name: lag_days
+      required: false
+      description: |-
+        How many days behind today to poll. Anthropic publishes daily aggregates
+        with a roughly 3-day lag; the default reflects that. Increase if you see
+        empty responses for the default date.
+      value:
+        type: integer
+        example: 3
+    - template: instances/default