Agent Academy is a free, source-linked library of skills, prompts, MCP servers and bot templates, organised into courses by profession, for people who hand real work to AI agents: 1,467 live resources, each with its original link, license and safety tags.
MCP server
Original text · EnglishA [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, companies, agents, groups, the knowledge base, and SLA/business-hours configuratio
https://github.com/WYRE-AI/freshdesk-mcp
MCP Registry: io.github.WYRE-AI/freshdesk-mcp; needs Freshdesk API keyRepublished in full under its license, with attribution. The source is authoritative.
A Model Context Protocol (MCP) server that provides AI assistants with structured access to the Freshdesk customer support platform — tickets, contacts, companies, agents, groups, the knowledge base, and SLA/business-hours configuration.
> Note: This project is maintained by Wyre Technology. It wraps the published @wyre-technology/node-freshdesk SDK.
The primary deployment is through the WYRE MCP Gateway, which hosts the containerized server and injects per-request credentials. To run it yourself:
Claude Code (CLI):
claude mcp add freshdesk-mcp \
-e FRESHDESK_DOMAIN=your-subdomain \
-e FRESHDESK_API_KEY=your-api-key \
-- npx -y github:WYRE-AI/freshdesk-mcp
FRESHDESK_DOMAIN is the part before .freshdesk.com (for https://acme.freshdesk.com it is acme). Find your FRESHDESK_API_KEY in the Freshdesk portal under Profile Settings.
freshdesk_navigate, enter a domain to reveal its tools, and freshdesk_back to return — keeping the tool list small and focusedWhen AUTH_MODE=gateway, every /mcp request must include credentials injected by the WYRE MCP Gateway. A request missing either header returns 401 and does not fall back to FRESHDESK_DOMAIN or FRESHDESK_API_KEY in the process environment.
| Header | Maps to | Description |
|---|---|---|
x-freshdesk-domain | FRESHDESK_DOMAIN | Freshdesk account subdomain |
x-freshdesk-api-key | FRESHDESK_API_KEY | Freshdesk API key |
Each request is stateless: the gateway provides the credentials, and the server rebuilds its Freshdesk client for that request only. When CONDUIT_S2S_SECRET is set, the request must also include a valid X-Gateway-S2S header.
Copy [.env.example](.env.example) and set the variables below. The stdio transport reads FRESHDESK_DOMAIN and FRESHDESK_API_KEY and does not use CONDUIT_S2S_SECRET.
| Variable | Description | Default |
|---|---|---|
FRESHDESK_DOMAIN | Account subdomain (before .freshdesk.com). Required for stdio and for HTTP when AUTH_MODE=env. Gateway mode does not read it. | — |
FRESHDESK_API_KEY | Freshdesk API key. Required for stdio and for HTTP when AUTH_MODE=env. Gateway mode does not read it. | — |
MCP_TRANSPORT | Transport mode (stdio or http) | stdio |
AUTH_MODE | env (process environment) or gateway (per-request x-freshdesk-domain and x-freshdesk-api-key, no env fallback). The container image and Compose default to gateway. | env |
CONDUIT_S2S_SECRET | Required for HTTP. HMAC secret checked against X-Gateway-S2S on every /mcp request. The process logs an error and exits non-zero when this is empty or a known placeholder (changeme, placeholder, secret, your-s2s-secret, and similar public example values). Compose will not start when it is unset. Never commit a real value. Generate one with openssl rand -hex 32. The value is never logged. | — |
MCP_ALLOW_INSECURE_DEV | Set to exactly 1 only for local development to start HTTP without CONDUIT_S2S_SECRET. The process logs a warning, does not enforce X-Gateway-S2S, and binds 127.0.0.1 only. A wider MCP_HTTP_HOST (including 0.0.0.0) is ignored and the override is logged. A known placeholder secret is still rejected. | — |
MCP_HTTP_HOST | Bind address when MCP_TRANSPORT=http. Unset binds 127.0.0.1. A configured CONDUIT_S2S_SECRET honors this value (the container image sets 0.0.0.0). With no secret, the insecure-dev bypass forces 127.0.0.1. | 127.0.0.1 |
MCP_HTTP_PORT | HTTP server port | 8080 |
LOG_LEVEL | Log level (debug, info, warn, error) | info |
/health stays unauthenticated and does not read credentials.
With MCP_TRANSPORT=http and CONDUIT_S2S_SECRET set, every /mcp request must include X-Gateway-S2S:
t=<unix seconds>,v1=<hex hmac-sha256>
v1 is the lowercase hex HMAC-SHA256 of the ASCII message t=<unix seconds>, keyed by CONDUIT_S2S_SECRET. The timestamp must be within 300 seconds of the server clock. A missing or invalid header returns 401 before any vendor credential is read. The secret is never written to logs or error bodies.
In AUTH_MODE=gateway, /mcp also requires x-freshdesk-domain and x-freshdesk-api-key. Those headers are the only credential source for the request. A missing header returns 401. The server does not fall back to FRESHDESK_DOMAIN or FRESHDESK_API_KEY in the process environment.
X-Gateway-S2S: t=1710000000,v1=<hmac-sha256 of "t=1710000000">
x-freshdesk-domain: acme
x-freshdesk-api-key:
Single-tenant HTTP (AUTH_MODE=env) still requires a valid X-Gateway-S2S header. Vendor credentials then come from FRESHDESK_DOMAIN and FRESHDESK_API_KEY.
The server uses decision-tree navigation. At the top level you see only freshdesk_navigate, freshdesk_back, and freshdesk_status. Call freshdesk_navigate with a domain to reveal that domain's tools; call freshdesk_back to return to the top.
| Domain | Tools |
|---|---|
| tickets | search, list, get, create, update, reply, add_note, list_conversations, update_conversation, delete\*, delete_conversation\* |
| contacts | search, list, get, create, update, autocomplete, make_agent†, restore, send_invite, merge†, soft_delete†, hard_delete\* |
| companies | search, list, get, create, update, autocomplete, delete\* |
| agents | list, get, me, create, update, delete\* |
| groups | list, get, create, update, delete\* |
| solutions | categories / folders / articles: list, get, create, update, delete\* |
| sla-business | SLA policies (list, create, update), business hours (list, get), canned responses (list_folders, get_folder, list_responses) |
* = Tier A (irreversible delete) — flagged ⚠ DESTRUCTIVE — IRREVERSIBLE, requires confirmation.
† = Tier B (high-impact reversible) — flagged ⚠ HIGH-IMPACT, requires confirmation.
Read tools (search/list/get/autocomplete/status) carry no warning and are marked read-only.
freshdesk_status performs a live connectivity check by calling agents/me.
The image listens on 0.0.0.0 inside the container (MCP_HTTP_HOST) and defaults to AUTH_MODE=gateway. Publish the port on loopback and pass the S2S secret. In gateway mode the domain and API key arrive on each request as x-freshdesk-domain and x-freshdesk-api-key. They are not read from the environment.
docker run \
-e CONDUIT_S2S_SECRET="$CONDUIT_S2S_SECRET" \
-p 127.0.0.1:8080:8080 \
freshdesk-mcp
Compose refuses to start when CONDUIT_S2S_SECRET is unset, sets AUTH_MODE=gateway, publishes 127.0.0.1:8080:8080, and does not set FRESHDESK_DOMAIN, FRESHDESK_API_KEY, or env_file:
CONDUIT_S2S_SECRET="$CONDUIT_S2S_SECRET" docker compose up -d
Single-tenant HTTP (AUTH_MODE=env) still requires the S2S secret, then uses FRESHDESK_DOMAIN and FRESHDESK_API_KEY:
docker run \
-e CONDUIT_S2S_SECRET="$CONDUIT_S2S_SECRET" \
-e AUTH_MODE=env \
-e FRESHDESK_DOMAIN=your-subdomain \
-e FRESHDESK_API_KEY=your-api-key \
-p 127.0.0.1:8080:8080 \
freshdesk-mcp
Local HTTP without a secret is development-only. It binds 127.0.0.1 even when MCP_HTTP_HOST is 0.0.0.0, and logs a warning. A known placeholder secret still refuses to start:
MCP_ALLOW_INSECURE_DEV=1 npm run start:http
See [docker-compose.yml](docker-compose.yml). The build installs @wyre-technology/node-freshdesk from GitHub Packages via a
BuildKit npmrc secret (never baked into a layer):
printf '@wyre-ai:registry=https://npm.pkg.github.com\n@wyre-technology:registry=https://npm.pkg.github.com\n//npm.pkg.github.com/:_authToken=%s\n' "$(gh auth token)" > /tmp/.npmrc
docker build --secret id=npmrc,src=/tmp/.npmrc -t freshdesk-mcp .
rm /tmp/.npmrc
export NODE_AUTH_TOKEN=$(gh auth token) # to install @wyre-technology/node-freshdesk
npm install
npm run build # Build the project
npm run dev # Watch mode
npm run test # Run tests
npm run lint # Type-check
npm run clean # Remove dist/
# HTTP refuses to start until CONDUIT_S2S_SECRET is set
CONDUIT_S2S_SECRET="$CONDUIT_S2S_SECRET" npm run start:http
# Local HTTP only, without a service-to-service secret (always binds 127.0.0.1)
MCP_ALLOW_INSECURE_DEV=1 npm run start:http
npm test # Run test suite
npm run test:watch # Watch mode
See [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines.
Apache 2.0 — Copyright WYRE Technology
A structured output agent demonstrating Pydantic schema-based responses for customer support ticket creation.
Sourcegithub.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
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 priority levels, categories, and resolution estimates using Pydantic schemas and
Sourcegithub.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
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 Qdrant as the vector store.
Sourcegithub.com/Shubhamsaboo/awesome-llm-apps/tree/main/advanced_ai_agents/single_agent_apps/ai_customer_support_agent
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 determinism to how you build your wrapping application, as you can count on a to
Sourcegithub.com/openai/openai-cookbook/blob/main/examples/Using_tool_required_for_customer_service.ipynb
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 information, retrieve order details, and cancel orders on behalf of the customer. We
Sourcegithub.com/anthropics/claude-cookbooks/blob/main/tool_use/customer_service_agent.ipynb
Curated collection of AI agent use cases across industries (incl. Customer Service, E-commerce) with links to open-source implementations, e.g. a LangGraph customer-support agent tutorial.
Sourcegithub.com/ashishpatel26/500-AI-Agents-Projects