setup-agent-analytics — плагин для Claude Code: как установить
Находит ИИ-агентов в кодовой базе и настраивает Pendo Agent Analytics через Python SDK, TypeScript SDK, вызов trackAgent() или Conversations API.
/plugin install setup-agent-analytics@claude-plugins-official
/plugin marketplace add anthropics/claude-plugins-officialЧто делает setup-agent-analytics
Плагин находит реализации ИИ-агентов в проекте и помогает внедрить Pendo Agent Analytics. Для интеграции подбирается подходящий способ под стек: серверный Python SDK, TypeScript SDK, клиентский метод trackAgent() или прямое обращение к Conversations API.
Инструмент рассчитан на работу с существующей агентной логикой. Если в кодовой базе нет кода агентов или подходящего окружения, плагину будет некуда привязать сбор телеметрии.
Какие задачи решает
- найти участки с ИИ-агентами в кодовой базе
- внедрить Pendo Agent Analytics через Python SDK или TypeScript SDK
- настроить сбор метрик агента на клиенте с помощью
trackAgent() - подключить аналитику диалогов через Conversations API
Примеры запросов агенту
- Найди ИИ-агентов в проекте и настрой для них Pendo Agent Analytics
- Подключи Pendo Agent Analytics к моим агентам через TypeScript SDK
- Добавь клиентский трекинг
trackAgent()для агентских сценариев
Как установить setup-agent-analytics
Через маркетплейс плагинов
/plugin install setup-agent-analytics@claude-plugins-official/plugin marketplace add anthropics/claude-plugins-officialПо шагам, с командами для каждой платформы: Плагины Claude Code: маркетплейсы, установка и как собрать свой
Вопросы о setup-agent-analytics
Как установить setup-agent-analytics в Claude Code?
Сначала добавьте каталог официальных плагинов: /plugin marketplace add anthropics/claude-plugins-official. Затем выполните установку: /plugin install setup-agent-analytics@claude-plugins-official.
Что делает setup-agent-analytics?
Он сканирует проект, обнаруживает логику ИИ-агентов и внедряет аналитику Pendo наиболее подходящим способом для вашей кодовой базы.
Какие способы интеграции поддерживает плагин?
Поддерживаются четыре пути: Python SDK, TypeScript SDK, клиентский вызов trackAgent() и Conversations API.
Исходный текст
setup-agent-analytics · MIT
# Pendo plugins for Claude Code Pendo analytics for Claude Code: app health, account health, feature adoption, session replays, feedback analysis, data-informed planning, agent analytics setup, and Orchestrate journey building. ## Plugins This marketplace contains five plugins: | Plugin | Description | |:-------|:------------| | `pendo-analytics` | Pendo analytics skills for account health, app health, feature adoption, session replays, feedback analysis, and data-informed planning | | `setup-agent-analytics` | Detect AI agents in your codebase and instrument them with Pendo agent analytics | | `setup-mcp-agent-analytics` | Detect an MCP server's language (Python, TypeScript, Go, or Java) and instrument it with the matching Pendo SDK for MCP analytics | | `pendo-guides` | Create production-ready Pendo in-app guides (walkthroughs, announcements, alerts, polls, promotions) as HTML/CSS/JS from a short intake conversation, and review guide signals to mute, deactivate, or archive stale guides | | `pendo-orchestrate` | Create, configure, and edit draft Orchestrate email journeys via Pendo MCP — multi-email journeys, conditional splits, and email content; activation in Orchestrate UI | ## Quickstart ### Option A: Install from Marketplace (Recommended) 1. **Add the marketplace:** Inside Claude Code, run: ``` /plugin marketplace add pendo-io/claude-pendo-plugin ``` 2. **Install plugins:** Run `/plugin`, navigate to the **Marketplaces** section, select the **pendo-plugins** marketplace, and install the plugins you want. 3. **Authenticate the Pendo MCP server:** Run the `/mcp` command inside Claude Code and follow the authentication flow. 4. **Run a skill:** **pendo-analytics:** ``` /pendo-analytics:app-health <app-name> /pendo-analytics:account-health <account-name> /pendo-analytics:feature-adoption <feature-name> /pendo-analytics:feedback-analysis /pendo-analytics:data-informed-planning <planning question> /pendo-analytics:replay-triage <ticket ID or replay-url> ``` **setup-agent-analytics:** ``` /setup-agent-analytics:setup-agent-analytics [agent-id-1] [agent-id-2] ... ``` **setup-mcp-agent-analytics:** ``` /setup-mcp-agent-analytics:setup-mcp-agent-analytics ``` **pendo-guides:** ``` /pendo-guides:pendo-guide-creator /pendo-guides:guide-management ``` **pendo-orchestrate:** ``` /pendo-orchestrate:orchestrate-journeys ``` ### Option B: Manual Installation 1. **Clone the repo:** ```bash git clone https://github.com/pendo-io/claude-pendo-plugin.git ``` 2. **Start Claude Code with the plugin:** ```bash claude --plugin-dir /path/to/claude-pendo-plugin ``` 3. **Authenticate the Pendo MCP server:** Run the `/mcp` command inside Claude Code and follow the authentication flow. 4. **Run a skill:** **pendo-analytics:** ``` /pendo-analytics:app-health <app-name> /pendo-analytics:account-health <account-name> /pendo-analytics:feature-adoption <feature-name> /pendo-analytics:feedback-analysis /pendo-analytics:data-informed-planning <planning question> /pendo-analytics:replay-triage <ticket ID or replay-url> ``` **setup-agent-analytics:** ``` /setup-agent-analytics:setup-agent-analytics [agent-id-1] [agent-id-2] ... ``` **setup-mcp-agent-analytics:** ``` /setup-mcp-agent-analytics:setup-mcp-agent-analytics ``` **pendo-guides:** ``` /pendo-guides:pendo-guide-creator /pendo-guides:guide-management ``` **pendo-orchestrate:** ``` /pendo-orchestrate:orchestrate-journeys ``` ## Components ### pendo-analytics skills | Skill | Description | |:------|:------------| | `app-health` | Generate a shareable health report for one application — engagement, acquisition, retention, average time, stickiness, frustration signals, and survey scores, with a verdict and next steps | | `account-health` | Prepare for a customer call by synthesizing engagement, sentiment, and feedback from Pendo analytics | | `feature-adoption` | Analyze feature adoption rates, identify power users vs laggards, and track adoption trends | | `feedback-analysis` | Deep analysis of customer feedback - discover themes, extract insights, and identify risks | | `data-informed-planning` | Ground a planning task in real product data — gathers signals, issues, wiki, customer feedback, PES/NPS, segments, and session replays from Novus + Pendo, then delegates to `/superpowers:writing-plans` | | `replay-triage` | Triage errors from a session replay — pull devlogs, locate the bug in the codebase, propose a fix, and prompt for a PR | ### setup-agent-analytics skills | Skill | Description | |:------|:------------| | `setup-agent-analytics` | Detect AI agents in a codebase and instrument them with Pendo `trackAgent()` calls or the server-side Conversations API | ### setup-mcp-agent-analytics skills | Skill | Description | |:------|:------------| | `setup-mcp-agent-analytics` | Detect an MCP server's language (Python, TypeScript, Go, or Java) and instrument it with the matching Pendo SDK (`PendoMCPServer`, `initMcp()`, `gosdk.Instrument()`, or `PendoMcpServer.wrapAll()`) | ### pendo-guides skills | Skill | Description | |:------|:------------| | `pendo-guide-creator` | Runs a short intake conversation, then generates production-ready Pendo in-app guides (walkthroughs, announcements, alerts, polls, promotions) as standalone HTML/CSS/JS files with wired button actions and `<pendo-poll>` data collection | | `guide-management` | Review the weekly Guide Signals for an app — archive, deactivate, and anomaly recommendations with their evidence and as-of date — then mute a signal or deactivate/archive the guide on your explicit choice | ### pendo-orchestrate skills | Skill | Description | |:------|:------------| | `orchestrate-journeys` | Lifecycle skill for draft Orchestrate journeys: intake, create, configure audience/schedule/goal, write email HTML, and hand off for activation in the Orchestrate UI | ### MCP Servers The `pendo-analytics`, `pendo-guides`, and `pendo-orchestrate` plugins auto-configure the Pendo connector (`pendo-external`). The `pendo-analytics` plugin also configures Novus: | Server | URL | Purpose | |:-------|:----|:--------| | `pendo-external` | `https://app.pendo.io/mcp/v0/shttp` | Pendo analytics, feedback, session replays, segments | | `novus` | `https://novus-api.pendo.io/mcp` | Pendo's Novus product-intelligence beta — analyzes a connected repo and proposes instrumentation changes as PRs — exposed over MCP | Both servers use the same Pendo OAuth login and are covered by the same [Pendo Terms of Service](https://www.pendo.io/legal/terms-of-service/) and [Privacy Policy](https://www.pendo.io/legal/privacy-policy/). The `novus` server is additionally covered by the Novus beta terms. Run `/mcp` once after installing the plugin to authenticate each server. ### Pendo MCP tools (`pendo-external`) | Tool | Purpose | |:-----|:--------| | `activityQuery` | Engagement metrics and activity data | | `productEngagementScore` | PES calculations | | `appUsage` | App-level active visitors/accounts, events, average time, and frustration counts (rage/dead/error clicks, u-turns) | | `appUsageTimeSeries` | Per-period app usage trend; basis for DAU/MAU stickiness | | `acquisitionTrend` | New visitors and accounts acquired per period | | `cohortRetentionCurve` | Retention curve for a cohort of new visitors | | `surveyScores` | NPS, CSAT, and PMF survey scores with response counts | | `searchEntities` | Find accounts, pages, features | | `accountQuery` | Account metadata | | `accountMetadataSchema` | Account metadata schema | | `visitorQuery` | Visitor metadata | | `sessionReplayList` | Find session recordings | | `devlogEvents` | Console errors and HTTP activity captured during a session replay | | `generate_feedback_topics` | Cluster feedback into themes | | `get_feedback_insights` | Extract key insights | | `get_feedback_items` | Raw feedback data | | `get_ideas` | Voice of Customer ideas / feature requests | | `guideMetrics` | Guide performance metrics | | `npsScore` | NPS scores (subscription must include NPS) | | `segmentList` | Available segments | | `list_all_applications` | List Pendo applications | ### Novus MCP tools (`novus`) | Tool | Purpose | |:-----|:--------| | `get_pendo_apps` | List Novus-tracked apps | | `get_product_wiki` | Auto-generated product wiki / sitemap | | `list_signals` | AI-generated product signals (issues, opportunities, trends) | | `list_issues` | AI-detected issues with recommended actions | | `list_artifacts` / `get_artifact` | Artifact graph nodes (pages, features, funnels, journeys, etc.) | | `get_related_artifacts` / `get_artifact_subtree` | Artifact relationships | | `get_page_metrics` / `get_feature_metrics` | Pre-processed engagement metrics | | `get_funnel_analysis` / `get_journey_analysis` | Conversion path analysis | | `get_headline_metrics` | Top-line product health | | `preview_guide` | Render a Pendo guide preview | | `create_flag` / `update_flag` | Feature flag tooling | ### Recommended Plugins The `data-informed-planning` skill can optionally delegate plan-writing to `/superpowers:writing-plans` for richer output. If the [superpowers plugin](https://github.com/obra/superpowers) is installed, the skill will use it automatically; otherwise it writes the plan itself. To install superpowers: ``` /plugin marketplace add obra/superpowers-marketplace ``` ## Privacy Use of the Pendo MCP servers is subject to the [Pendo Privacy Policy](https://www.pendo.io/legal/privacy-policy/) and [Terms of Service](https://www.pendo.io/legal/terms-of-service/). ## License MIT
© Pendo · MIT · pendo-io/claude-pendo-plugin