---
name: customer-service-customer-success
description: Agent Academy course for Customer success (Customer service). Five ordered steps of real, source-linked resources.
version: 2026-10-15
updated_at: 2026-10-15T10:41:56Z
source: https://siinxu.github.io/agent-academy/courses/customer-service/customer-success/
---

# Customer success · Customer service

Goal: Learn to triage, answer and escalate customer requests with written policies, a help-desk MCP if allowed, and a self-check before sending.

Permissions: Read-only by default. Help-desk MCP servers (Zendesk, Freshdesk, Intercom, ...) need API keys and can send replies; ask the owner first.

Machine-readable version: https://raw.githubusercontent.com/SiinXu/agent-academy/main/academy/courses/customer-service/customer-success.json
Skill pack: https://siinxu.github.io/agent-academy/packs/customer-service/customer-success.zip
Changes feed: https://siinxu.github.io/agent-academy/api/changes.json

Rules for the agent: read the steps in order. For each item open `raw` if present (mirrored text, license allows it), otherwise open `source`. Do not install or use anything marked needs_credentials or writes_external without asking your owner.


## Step 1: Enrol

Read course.md: the job, the permissions it needs, and what good output looks like.

- [tutorial] 2_1_support_ticket_agent. A structured output agent demonstrating Pydantic schema-based responses for customer support ticket creation.
  source: https://github.com/Shubhamsaboo/awesome-llm-apps/tree/main/ai_agent_framework_crash_course/openai_sdk_crash_course/2_structured_output_agent/2_1_support_ticket_agent
  raw: https://raw.githubusercontent.com/SiinXu/agent-academy/main/data/mirror/shubhamsaboo-awesome-llm-apps-ai-agent-framework-crash-course-openai-sdk-crash-course-2-st.md
  license: Apache-2.0
  needs_credentials: false, writes_external: false
- [tutorial] 3_1_customer_support_ticket_agent. A tutorial demonstrating how to implement a structured customer support ticketing system using Google's ADK (Agent Development Kit) framework. This example shows how to create type-safe, structured support tickets with p
  source: https://github.com/Shubhamsaboo/awesome-llm-apps/tree/main/ai_agent_framework_crash_course/google_adk_crash_course/3_structured_output_agent/3_1_customer_support_ticket_agent
  raw: https://raw.githubusercontent.com/SiinXu/agent-academy/main/data/mirror/shubhamsaboo-awesome-llm-apps-ai-agent-framework-crash-course-google-adk-crash-course-3-st.md
  license: Apache-2.0
  needs_credentials: false, writes_external: false
- [tutorial] Using tool_choice=required for customer service (OpenAI cookbook). Using Tool Required for Customer Service The ChatCompletion endpoint now includes the ability to specify whether a tool must be called every time, by adding tool_choice='required' as a parameter. This adds an element of 
  source: https://github.com/openai/openai-cookbook/blob/main/examples/Using_tool_required_for_customer_service.ipynb
  raw: https://raw.githubusercontent.com/SiinXu/agent-academy/main/data/originals/openai-openai-cookbook-examples-using-tool-required-for-customer-service-ipynb.md
  license: MIT
  needs_credentials: false, writes_external: false
- [tutorial] Customer service agent with tool use (Claude cookbook). Creating a Customer Service Agent with Client-Side Tools In this recipe, we'll demonstrate how to create a customer service chatbot using Claude 3 plus client-side tools. The chatbot will be able to look up customer info
  source: https://github.com/anthropics/claude-cookbooks/blob/main/tool_use/customer_service_agent.ipynb
  raw: https://raw.githubusercontent.com/SiinXu/agent-academy/main/data/originals/anthropics-claude-cookbooks-tool-use-customer-service-agent-ipynb.md
  license: MIT
  needs_credentials: false, writes_external: false

## Step 2: Pick a direction

Choose one style, platform or channel from the reference library. Only one.

- [template] Big Tony. Your enforcer against AI customer service. Big Tony handles cancellations that weren't honored, ghost charges, retention traps, fraud cleanup, and
  source: https://botsdots.site/templates/big-tony
  license: template belongs to its creator (catalog CC0)
  install: https://x.ai/bot/hBo0iWrkgWTueZe1TyGhm
  needs_credentials: false, writes_external: false
- [template] Churn retention manager. Steps in on cancellations and failed payments before you lose the customer
  source: https://botsdots.site/templates/churn-retention-manager
  license: template belongs to its creator (catalog CC0)
  install: https://x.ai/bot/p0OIZUJK5sRFImh_DMujf
  needs_credentials: false, writes_external: true
- [plugin] Customer Support plugin (Anthropic knowledge-work-plugins). A customer support plugin primarily designed for [Cowork](https://claude.com/product/cowork), Anthropic's agentic desktop application — though it also works in Claude Code. Provides ticket triage, escalation management, 
  source: https://github.com/anthropics/knowledge-work-plugins/tree/main/customer-support
  raw: https://raw.githubusercontent.com/SiinXu/agent-academy/main/data/mirror/anthropics-knowledge-work-plugins-customer-support.md
  license: Apache-2.0
  install: claude plugin marketplace add anthropics/knowledge-work-plugins
  needs_credentials: false, writes_external: false
- [plugin] ADK retail plugin. The **Retail AI Plugin** bundles two end-to-end Google Cloud retail skills for AI coding assistants (Gemini CLI, Claude Code, Antigravity, Codex) following the [Google Agent Plugins v1.0.0 specification][spec] (`plugin.j
  source: https://github.com/google/adk-recipes/tree/main/plugins/retail
  raw: https://raw.githubusercontent.com/SiinXu/agent-academy/main/data/originals/google-adk-recipes-plugins-retail.md
  license: Apache-2.0
  needs_credentials: false, writes_external: false

## Step 3: Learn the rules

Read the required skills, prompts and standards in order.

- [skill] churn-prevention. When the user wants to reduce churn, build cancellation flows, set up save offers, recover failed payments, or implement retention strategies. Also use when the user mentions 'churn,' 'cancel flow,' 'offboarding,' 'save 
  source: https://github.com/coreyhaines31/marketingskills/blob/main/skills/churn-prevention/SKILL.md
  raw: https://raw.githubusercontent.com/SiinXu/agent-academy/main/data/originals/coreyhaines31-marketingskills-skills-churn-prevention-skill-md.md
  license: MIT
  install: npx skills add coreyhaines31/marketingskills --skill churn-prevention
  needs_credentials: false, writes_external: false
- [system-prompt] customer-success-manager. Use this agent when you need to assess customer health, develop retention strategies, identify upsell opportunities, or maximize customer lifetime value. Invoke this agent for account health analysis, churn prevention, p
  source: https://github.com/VoltAgent/awesome-claude-code-subagents/blob/main/categories/08-business-product/customer-success-manager.md
  raw: https://raw.githubusercontent.com/SiinXu/agent-academy/main/data/mirror/voltagent-awesome-claude-code-subagents-categories-08-business-product-customer-success-ma.md
  license: MIT
  needs_credentials: false, writes_external: false
- [skill] customer-escalation. Package an escalation for engineering, product, or leadership with full context. Use when a bug needs engineering attention beyond normal support, multiple customers report the same issue, a customer is threatening to ch
  source: https://github.com/anthropics/knowledge-work-plugins/blob/main/customer-support/skills/customer-escalation/SKILL.md
  raw: https://raw.githubusercontent.com/SiinXu/agent-academy/main/data/originals/anthropics-knowledge-work-plugins-customer-support-skills-customer-escalation-skill-md.md
  license: Apache-2.0
  install: npx skills add anthropics/knowledge-work-plugins --skill customer-escalation
  needs_credentials: false, writes_external: false
- [skill] customer-research. Multi-source research on a customer question or topic with source attribution. Use when a customer asks something you need to look up, investigating whether a bug has been reported before, checking what was previously to
  source: https://github.com/anthropics/knowledge-work-plugins/blob/main/customer-support/skills/customer-research/SKILL.md
  raw: https://raw.githubusercontent.com/SiinXu/agent-academy/main/data/originals/anthropics-knowledge-work-plugins-customer-support-skills-customer-research-skill-md.md
  license: Apache-2.0
  install: npx skills add anthropics/knowledge-work-plugins --skill customer-research
  needs_credentials: false, writes_external: false

## Step 4: Do it

Follow a workflow or case. Install a skill or MCP server only if you have permission.

- [case] ai_customer_support_agent. This Streamlit app implements an AI-powered customer support agent for synthetic data generated using GPT-4o. The agent uses OpenAI's GPT-4o model and maintains a memory of past interactions using the Mem0 library with Q
  source: https://github.com/Shubhamsaboo/awesome-llm-apps/tree/main/advanced_ai_agents/single_agent_apps/ai_customer_support_agent
  raw: https://raw.githubusercontent.com/SiinXu/agent-academy/main/data/mirror/shubhamsaboo-awesome-llm-apps-advanced-ai-agents-single-agent-apps-ai-customer-support-age.md
  license: Apache-2.0
  needs_credentials: false, writes_external: false
- [case] OpenAI customer service agents demo (Agents SDK). This repository contains a demo of a Customer Service interface built on top of the [OpenAI Agents SDK](https://openai.github.io/openai-agents-python/).
  source: https://github.com/openai/openai-cs-agents-demo
  raw: https://raw.githubusercontent.com/SiinXu/agent-academy/main/data/originals/openai-openai-cs-agents-demo.md
  license: MIT
  needs_credentials: false, writes_external: false
- [mcp] Zendesk MCP server. A Model Context Protocol server for Zendesk.
  source: https://github.com/reminia/zendesk-mcp-server
  raw: https://raw.githubusercontent.com/SiinXu/agent-academy/main/data/originals/reminia-zendesk-mcp-server.md
  license: Apache-2.0
  install: see README (uv / Claude Desktop config); needs Zendesk API token
  needs_credentials: true, writes_external: false
- [mcp] Freshdesk MCP server. A [Model Context Protocol (MCP)](https://modelcontextprotocol.io/) server that provides AI assistants with structured access to the [Freshdesk](https://www.freshdesk.com) customer support platform — tickets, contacts, co
  source: https://github.com/WYRE-AI/freshdesk-mcp
  raw: https://raw.githubusercontent.com/SiinXu/agent-academy/main/data/originals/wyre-ai-freshdesk-mcp.md
  license: Apache-2.0 (LICENSE file; GitHub API reports NOASSERTION)
  install: MCP Registry: io.github.WYRE-AI/freshdesk-mcp; needs Freshdesk API key
  needs_credentials: true, writes_external: false

## Step 5: Self-check

Go through the checklist and report the result to your owner.

- [benchmark] tau2-bench (customer-service agent benchmark). τ-bench family benchmark (MIT) that simulates user–agent conversations in customer-service domains (retail, airline, telecom, banking) with tools, policies and automatic task evaluation.
  source: https://github.com/sierra-research/tau2-bench
  raw: https://raw.githubusercontent.com/SiinXu/agent-academy/main/data/originals/sierra-research-tau2-bench.md
  license: MIT
  needs_credentials: false, writes_external: false
- [skill] customer-escalation. Package an escalation for engineering, product, or leadership with full context. Use when a bug needs engineering attention beyond normal support, multiple customers report the same issue, a customer is threatening to ch
  source: https://github.com/anthropics/knowledge-work-plugins/blob/main/customer-support/skills/customer-escalation/SKILL.md
  raw: https://raw.githubusercontent.com/SiinXu/agent-academy/main/data/originals/anthropics-knowledge-work-plugins-customer-support-skills-customer-escalation-skill-md.md
  license: Apache-2.0
  install: npx skills add anthropics/knowledge-work-plugins --skill customer-escalation
  needs_credentials: false, writes_external: false

### Checklist

- [ ] Every reply answers the customer's actual question in the first two sentences.
- [ ] Refunds, cancellations and account changes follow the written policy; anything outside it is escalated, not improvised.
- [ ] Tickets are tagged with category and priority (P1-P4) before they are answered.
- [ ] No customer data, order numbers or emails are pasted into third-party tools that are not approved.
- [ ] Sending a reply, closing a ticket or issuing a refund through an MCP server happens only after the owner approved that tool.
- [ ] Knowledge-base articles cite the ticket pattern they came from and are reviewed before publishing.
- [ ] Escalations include a one-paragraph summary, what was tried, and what the customer was told.


Report back to your owner: what you learned (5 bullets), which items you installed (none without approval), and the checklist results.