diff --git a/src/content/docs/enterprise/enterprise-features/analytics-api.mdx b/src/content/docs/enterprise/enterprise-features/analytics-api.mdx
new file mode 100644
index 0000000..2f0b7b8
--- /dev/null
+++ b/src/content/docs/enterprise/enterprise-features/analytics-api.mdx
@@ -0,0 +1,310 @@
+---
+title: Enterprise Analytics API
+description: >-
+  Programmatically access team-level usage data for Warp's enterprise plans —
+  per-team summaries, per-user rollups, and message-level activity events.
+sidebar:
+  label: "Analytics API"
+---
+
+The Enterprise Analytics API lets enterprise admins pull Warp usage data into their own dashboards, cost-allocation tooling, or audit pipelines. It exposes three read-only endpoints over HTTPS that return aggregated team metrics, per-user rollups, and message-level activity events for the agents your team runs in Warp and Oz.
+
+:::note[Early access]
+The Enterprise Analytics API is in Early Access. It is available to all enterprise teams. To start collecting data for your team, an admin must turn on **Enterprise Usage Reporting (Early Access)** in **Admin Panel** > **Privacy** — no usage data is recorded until that toggle is on.
+:::
+
+## What you can do with the API
+
+* **Build productivity dashboards** - Track adoption across users and teams over time using `summary` and `users`.
+* **Allocate credit spend** - Break down usage by user, model, or run type (`local` vs. `cloud`) to attribute cost to teams or projects.
+* **Audit message-level activity** - Filter the `events` endpoint by `run_type`, `user_id`, `conversation_id`, `message_type`, or `tool_type` for compliance reviews and incident investigation.
+* **Power custom reporting** - Pipe the JSON responses into your own warehouse, BI tool, or alerting system.
+
+## Prerequisites and access
+
+Before you can call the API, your team must satisfy all of the following:
+
+* **Enterprise plan** - The Analytics API is available to all enterprise teams during Early Access; no separate enrollment is required.
+* **Admin role on the team** - Calls are rejected unless the authenticated user has admin-level permissions on the enterprise team. See [Roles and permissions](/enterprise/team-management/roles-and-permissions/).
+* **A personal Warp API key** - Authenticate requests with a key from **Settings** > **Cloud platform** > **Oz Cloud API Keys** in the Warp app. See [API Keys](/reference/cli/api-keys/) for step-by-step instructions. Team API keys are not accepted by these endpoints — only personal API keys belonging to a team admin work.
+* **Enterprise Usage Reporting toggle enabled** - In the Warp app, go to **Admin Panel** > **Privacy** and turn on **Enterprise Usage Reporting (Early Access)**. Until this toggle is on, no usage data is recorded for your team and the endpoints will return empty datasets even if every other prerequisite is met.
+
+:::caution
+Personal API keys inherit the access of the user who created them. Treat them like any other admin credential and rotate them when team membership changes.
+:::
+
+## Quickstart
+
+The following example fetches a team-level summary for the last day. Set your API key first:
+
+```bash
+export WARP_API_KEY="wk-..."
+```
+
+Then call the `summary` endpoint:
+
+```bash
+curl "https://app.warp.dev/api/v1/enterprises/analytics/summary" \
+  -H "Authorization: Bearer $WARP_API_KEY"
+```
+
+Once an admin has turned on the **Enterprise Usage Reporting (Early Access)** toggle for your team, you'll get a JSON payload aggregating credit usage and conversation counts across the team. Continue to the [Endpoints](#endpoints) section for per-endpoint details.
+
+## Endpoints
+
+All endpoints share the same base URL, authentication, and shared validation rules:
+
+* **Base URL** - `https://app.warp.dev/api/v1/enterprises/analytics`
+* **Authentication** - `Authorization: Bearer $WARP_API_KEY` header on every request.
+* **Dates** - Pass `start_date` and `end_date` as RFC 3339 timestamps with timezone (for example, `2026-04-01T00:00:00Z` or `2026-04-01T00:00:00-07:00`). When omitted, the API defaults to a 1-day window ending at the current time. `start_date` must be strictly before `end_date`; otherwise the request returns `400 Bad Request`.
+
+### `GET /api/v1/enterprises/analytics/summary`
+
+Returns a single team-level rollup of conversation counts, credit spend, and code-change activity for the requested window. When `group_by_period` is set, the response is a time series with one bucket per period.
+
+**Query parameters:**
+
+* `start_date` (optional) - RFC 3339 timestamp. Defaults to one day before `end_date`.
+* `end_date` (optional) - RFC 3339 timestamp. Defaults to the current time.
+* `group_by_period` (optional) - One of `day`, `week`, or `month`. When set, the resulting `period_list` is capped at 365 periods; longer ranges with finer granularity are rejected with `400 Bad Request`.
+
+**Example:**
+
+```bash
+curl "https://app.warp.dev/api/v1/enterprises/analytics/summary?start_date=2026-04-01T00:00:00Z&end_date=2026-04-30T23:59:59Z&group_by_period=day" \
+  -H "Authorization: Bearer $WARP_API_KEY"
+```
+
+**Ungrouped response shape:**
+
+```json
+{
+  "data": {
+    "local": {
+      "total_conversations_count": 1234,
+      "total_credits_spent": 8421.5,
+      "code_changes": {
+        "file_changes": { "suggested": 412, "accepted": 305 },
+        "lines_of_code": {
+          "suggested": { "added_count": 9821, "removed_count": 2104 },
+          "accepted":  { "added_count": 7102, "removed_count": 1488 }
+        }
+      }
+    },
+    "cloud": {
+      "total_conversations_count": 412,
+      "total_credits_spent": 2310.4,
+      "code_changes": {
+        "file_changes": { "suggested": 88, "accepted": 64 },
+        "lines_of_code": {
+          "suggested": { "added_count": 1450, "removed_count": 320 },
+          "accepted":  { "added_count": 1102, "removed_count": 248 }
+        }
+      }
+    }
+  }
+}
+```
+
+**Grouped response shape (`group_by_period=day`):**
+
+```json
+{
+  "group_by_period": "day",
+  "period_list": ["2026-04-01", "2026-04-02", "..."],
+  "data": {
+    "local": {
+      "total_conversations_count": [42, 51, "..."],
+      "total_credits_spent": [310.1, 412.5, "..."],
+      "code_changes": {
+        "file_changes": { "suggested": [12, 18, "..."], "accepted": [9, 14, "..."] },
+        "lines_of_code": {
+          "suggested": { "added_count": [240, 380, "..."], "removed_count": [55, 90, "..."] },
+          "accepted":  { "added_count": [180, 290, "..."], "removed_count": [40, 70, "..."] }
+        }
+      }
+    },
+    "cloud": { "...": "same shape as local" }
+  }
+}
+```
+
+In the grouped response, every leaf array is the same length as `period_list`, and the index of each value corresponds to the same index in `period_list`.
+
+### `GET /api/v1/enterprises/analytics/users`
+
+Returns per-user aggregated usage, split into `local` (Warp app) and `cloud` (Oz cloud agents) sections. Results are paginated.
+
+**Query parameters:**
+
+* `start_date` (optional) - RFC 3339 timestamp. Defaults to one day before `end_date`.
+* `end_date` (optional) - RFC 3339 timestamp. Defaults to the current time.
+* `group_by_period` (optional) - One of `day`, `week`, or `month`. When set, each user record's per-run-type fields are arrays indexed by the response-level `period_list`.
+* `page` (optional) - Page number, 1-indexed. Default `1`.
+* `page_size` (optional) - Items per page. Default `10`.
+
+**Example:**
+
+```bash
+curl "https://app.warp.dev/api/v1/enterprises/analytics/users?page=1&page_size=25" \
+  -H "Authorization: Bearer $WARP_API_KEY"
+```
+
+**Ungrouped response shape:**
+
+```json
+{
+  "data": [
+    {
+      "user_id": "firebase-uid-abc",
+      "email": "alex@example.com",
+      "local": {
+        "distinct_conversation_count": 42,
+        "credits_spent": 310.1,
+        "model_usage": [
+          { "model_id": "claude-4-7-opus", "distinct_conversation_count": 30, "credits_spent": 240.0 }
+        ],
+        "code_changes": {
+          "file_changes": { "suggested": 18, "accepted": 12 },
+          "lines_of_code": {
+            "suggested": { "added_count": 412, "removed_count": 88 },
+            "accepted":  { "added_count": 305, "removed_count": 64 }
+          }
+        }
+      },
+      "cloud": { "...": "same shape as local" }
+    }
+  ],
+  "pagination": {
+    "current_page": 1,
+    "page_size": 25,
+    "total_pages": 4,
+    "total_count": 87
+  }
+}
+```
+
+When `group_by_period` is set, each `local`/`cloud` block's scalars become arrays indexed by the response-level `period_list` (same convention as the grouped `summary` response). Within `model_usage`, each entry's `distinct_conversation_count` and `credits_spent` likewise become arrays.
+
+### `GET /api/v1/enterprises/analytics/events`
+
+Returns paginated, message-level activity events for fine-grained audits and dashboards. Each row corresponds to a single message produced or processed by an agent — the same canonical message stream the client and Warp UI see.
+
+**Query parameters:**
+
+* `start_date` (optional) - RFC 3339 timestamp. Defaults to one day before `end_date`. The window between `start_date` and `end_date` cannot exceed 365 days.
+* `end_date` (optional) - RFC 3339 timestamp. Defaults to the current time.
+* `page` (optional) - Page number, 1-indexed. Default `1`.
+* `page_size` (optional) - Items per page. Default `10`. Maximum `100`.
+* `run_type` (optional) - Filter by `local` or `cloud`.
+* `user_id` (optional) - Firebase UID of a single team member.
+* `conversation_id` (optional) - Restrict results to one conversation.
+* `message_type` (optional) - Filter by message variant. Common values include `agent_output`, `tool_call`, and `tool_call_result`. The full set mirrors the `Message.message` proto oneof, so additional values may appear over time. Internal scaffolding messages (for example `model_used`, `server_event`, `messages_received_from_agents`, `events_from_agents`, `debug_output`) are filtered out at write time and never appear here.
+* `tool_type` (optional) - For `tool_call` / `tool_call_result` messages, restrict to a specific tool. Values are the snake_case proto field names of the tool (for example `apply_file_diffs`, `run_shell_command`, `search_codebase`, `grep`, `call_mcp_tool`).
+
+Events are returned newest-first, ordered by `message_end_timestamp DESC, message_id DESC`.
+
+**Example:**
+
+```bash
+curl "https://app.warp.dev/api/v1/enterprises/analytics/events?run_type=cloud&message_type=tool_call&tool_type=apply_file_diffs&page_size=100" \
+  -H "Authorization: Bearer $WARP_API_KEY"
+```
+
+**Response shape:**
+
+```json
+{
+  "data": [
+    {
+      "message_id": "msg-uuid",
+      "llm_request_ids": ["llm-req-uuid"],
+      "message_type": "tool_call",
+      "tool_type": "apply_file_diffs",
+      "message_start_timestamp": "2026-04-15T19:21:08Z",
+      "message_end_timestamp": "2026-04-15T19:21:11Z",
+      "conversation_id": "conv-uuid",
+      "user_id": "firebase-uid-abc",
+      "email": "alex@example.com",
+      "user_selected_model": "auto",
+      "run_type": "cloud",
+      "model_ids": ["claude-4-7-opus"],
+      "credit_charged": 0.42,
+      "code_changes": {
+        "file_changes": { "suggested": 3, "accepted": 2 },
+        "lines_of_code": {
+          "suggested": { "added_count": 48, "removed_count": 12 },
+          "accepted":  { "added_count": 36, "removed_count": 8 }
+        },
+        "was_edited_by_user": false
+      },
+      "git_context": {
+        "head": "a1b2c3d",
+        "branch": "feature/analytics-api"
+      },
+      "error_type": null
+    }
+  ],
+  "pagination": {
+    "current_page": 1,
+    "page_size": 100,
+    "total_pages": 12,
+    "total_count": 1180
+  }
+}
+```
+
+Fields are populated as follows:
+
+* `llm_request_ids` and `model_ids` are arrays. Most LLM-backed messages have one entry; multi-partition summarization can list several.
+* `credit_charged` is the per-message proportional share of the backing LLM request's cost. Messages with no LLM call (for example client-submitted `tool_call_result` rows) carry `0`.
+* `code_changes` is omitted (or `null`) for messages that don't suggest or apply file changes. When present, `was_edited_by_user` is `true` when the user edited the diff before accepting, `false` when accepted as-is, and `null` when the message had no acceptance event.
+* `git_context` is omitted (or `null`) when no git state was captured for the message.
+* `error_type` is populated only when the backing LLM request errored.
+
+## Pagination
+
+The `users` and `events` endpoints use offset-based pagination with `page` and `page_size` query parameters. Every paginated response includes a `pagination` block:
+
+* `current_page` - The page returned, 1-indexed.
+* `page_size` - The number of items requested per page.
+* `total_pages` - The total number of pages available for the current query.
+* `total_count` - The total number of records matching the query across all pages.
+
+To iterate, request `page=1` first, then increment `page` until you've fetched `total_pages` pages. The `events` endpoint caps `page_size` at `100`; values above that are rejected. The `users` endpoint defaults to `page_size=10`.
+
+## Common use cases
+
+* **Executive dashboards** - Combine the `summary` endpoint's grouped output with your BI tool to chart team-wide adoption and credit spend over time.
+* **Cost allocation** - Use the `users` endpoint to attribute credit spend to individuals, teams, or projects for chargeback or showback.
+* **Adoption monitoring** - Track per-user `distinct_conversation_count` over time to identify power users, onboard slow adopters, and measure rollout impact.
+* **Compliance audit trails** - Pull the `events` endpoint with `message_type=tool_call` filters to log code-changing actions for security review.
+
+## FAQ
+
+### How fresh is the data?
+
+Usage data is recorded as agents run and is typically available within a few minutes. There is no guaranteed SLA on freshness during Early Access.
+
+### Who can call the API?
+
+Any authenticated user with admin-level permissions on an enterprise team. Calls from non-admin users return `403 Forbidden`.
+
+### What kind of API key works?
+
+Only **personal** Warp API keys created by an admin from **Settings** > **Cloud platform** > **Oz Cloud API Keys**. Team API keys are explicitly rejected by these endpoints. See [API Keys](/reference/cli/api-keys/) for how to create one.
+
+### Are these calls billed?
+
+No. Analytics API calls don't consume Warp credits.
+
+### What time range can I query?
+
+The `events` endpoint enforces a hard 365-day window between `start_date` and `end_date`. The `summary` and `users` endpoints don't enforce a fixed window, but the grouped variants are capped at 365 periods.
+
+## Related resources
+
+* [API Keys](/reference/cli/api-keys/) - Create and manage personal Warp API keys.
+* [Admin panel](/enterprise/team-management/admin-panel/) - Manage team settings, including the **Privacy** section.
+* [Roles and permissions](/enterprise/team-management/roles-and-permissions/) - Required admin role for Analytics API access.
+* [Architecture and deployment](/enterprise/enterprise-features/architecture-and-deployment/) - Where enterprise data is stored and how it transits Warp's infrastructure.
diff --git a/src/sidebar.ts b/src/sidebar.ts
index dce0e15..c0cb72e 100644
--- a/src/sidebar.ts
+++ b/src/sidebar.ts
@@ -551,6 +551,7 @@ export const sidebarTopics: StarlightSidebarTopicsUserConfig = [
 					items: [
 						{ slug: 'enterprise/enterprise-features/architecture-and-deployment', label: 'Architecture and deployment' },
 						{ slug: 'enterprise/enterprise-features/bring-your-own-llm', label: 'Bring your own LLM' },
+						{ slug: 'enterprise/enterprise-features/analytics-api', label: 'Analytics API' },
 					],
 				},
 				{