Plug your AI agent into your workspace.

What is MCP?

Model Context Protocol (MCP) is an open standard that lets AI agents discover and call tools from external systems in a structured, typed way. Think of it as USB-C for AI — one port, many devices.

The Wendesk MCP server runs at mcp.wendesk.com. Once your agent connects (via OAuth), it can list and call 42 tools scoped to your tenant — read leads, draft replies, move deals across pipeline stages, send WhatsApp messages, generate content, schedule voice calls.

Quickstart.

Connect Claude Desktop to your Wendesk workspace in under 3 minutes.

1. Open the MCP install page

   In Wendesk: Settings → Developer → MCP server. You'll see a one-click install URL for each supported client.

2. Pick scopes

   Default scopes are crm:read, inbox:read, content:read. Toggle on write scopes (crm:write, wa:send) if you want the agent to take actions.

3. Click "Connect to Claude"

   Browser opens an OAuth consent screen → approve → redirected back to Claude with the connection live. Same flow for ChatGPT, Cursor, Copilot.

4. Try a query

   In Claude: "List my hottest qualified leads from this week." Claude introspects available tools, calls crm.lead.list, returns results.

OAuth setup (manual).

If you're building a custom MCP client (not Claude/ChatGPT/Cursor), here's the manual flow:

# 1. Generate code_verifier + code_challenge (PKCE)
code_verifier=$(openssl rand -base64 64 | tr -d "=+/" | cut -c1-128)
code_challenge=$(echo -n $code_verifier | openssl dgst -sha256 -binary | base64 | tr -d "=+/" | cut -c1-43)

# 2. Send user to authorize URL
AUTHORIZE_URL="https://mcp.wendesk.com/oauth/authorize?response_type=code&client_id=YOUR_CLIENT_ID&redirect_uri=YOUR_REDIRECT&scope=crm:read+inbox:read&code_challenge=$code_challenge&code_challenge_method=S256"

# 3. After user approves, exchange code for token
curl -X POST https://mcp.wendesk.com/oauth/token \
  -d "grant_type=authorization_code" \
  -d "code=AUTH_CODE_FROM_CALLBACK" \
  -d "redirect_uri=YOUR_REDIRECT" \
  -d "client_id=YOUR_CLIENT_ID" \
  -d "code_verifier=$code_verifier"

# Response — 200 OK
{
  "access_token": "wd_mcp_at_...",
  "refresh_token": "wd_mcp_rt_...",
  "token_type": "Bearer",
  "expires_in": 3600,
  "scope": "crm:read inbox:read"
}

Tool catalog.

42 tools across 5 namespaces. R = read-only, W = write, RW = both. Each requires the corresponding scope.

+ 29 more tools across analytics, billing, integrations, tenant, and industry namespaces. Full catalog at mcp.wendesk.com/tools.

Scopes & permissions.

14 scopes total. Apply principle-of-least-privilege — only request what your agent needs.

• crm:read · crm:write — leads, contacts, deals, activities
• inbox:read · inbox:write — conversations + messages across all channels
• wa:send — send templated WhatsApp messages
• content:read · content:write — Content Studio drafts + brand voices
• voice:schedule — schedule outbound voice agent calls
• analytics:read — aggregate metrics (no PII)
• integrations:read · integrations:write — App Store connections
• workflow:run — execute saved workflows
• tenant:read — workspace metadata + plan info

Connect Claude.

Two paths — Claude Desktop (stdio) and Claude.ai (web, SSE-based).

Claude Desktop config

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "wendesk": {
      "command": "npx",
      "args": ["-y", "@wendesk/mcp-server"],
      "env": {
        "WENDESK_TOKEN": "wd_mcp_at_...",
        "WENDESK_TENANT": "acmecorp"
      }
    }
  }
}

Restart Claude Desktop. The Wendesk tools appear in the 🔌 menu, ready to use.

Other clients.

Rate limits.

Per-token rate limits — generous, but bounded:

• Read tools — 600 calls/min, 30,000 calls/hour
• Write tools — 60 calls/min, 1,800 calls/hour
• Expensive tools (content.generate, voice.call.schedule) — 30 calls/min

Exceeded? You get 429 Too Many Requests with a Retry-After header. Wait that many seconds and try again.

Errors.

• 401 Unauthorized — token expired or revoked → refresh via OAuth refresh flow
• 403 Forbidden — token has insufficient scope for this tool → request additional scope and re-auth
• 404 Not Found — entity ID doesn't exist in this tenant
• 422 Unprocessable — input failed Zod validation; check the errors[] array in the response
• 429 Too Many Requests — rate-limited; wait and retry
• 500 Internal Server Error — log the trace ID and email [email protected]