For support agents
Get through the queue from your chat
Connect once, then ask your assistant to find, summarise, draft, tag and assign. It works on your real conversations, as you.
MCP server
Go from AI chat to closed ticket
ThriveDesk MCP lets Claude, ChatGPT, Cursor and VS Code search, reply, tag, assign and report on your real conversations, with your own sign-in and only the access you grant.
https://mcp.thrivedesk.com
Included on Standard and Unlimited · No API key
Claude ChatGPT
Cursor
VS Code Perplexity
ThriveDesk connected
Works with the AI tools you already use
- Claude
- Claude Code
- ChatGPT
- Cursor
- VS Code
- and any MCP client
Say what you need. The assistant finds the action.
Most help desk integrations make you set up each action ahead of time. With ThriveDesk MCP you ask for the outcome, and your assistant picks from 28 actions across conversations, contacts, live chat, reports and your knowledge base, then runs them as you.
How it works
Connected in a couple of minutes
Nothing to install, no API key and no developer. The connection uses your normal ThriveDesk sign-in, and you can take it back at any time.
- Step 1
Paste one address
Add https://mcp.thrivedesk.com to your assistant's connector settings. It is the same address for every customer, with no path and no key to fill in.
- Step 2
Approve it on ThriveDesk
A ThriveDesk page opens. You sign in there, not inside the assistant, pick the workspace, choose how much access to grant, and authorise.
- Step 3
Ask in plain English
Try “List my ThriveDesk inboxes.” If your real inbox names come back, you are done. From then on, every reply, note, and assignment lands in the thread under your name.
-
What comes next
Start with Read-only or drafts. Once you trust how your assistant works, let it send replies and run the queue. Everything it does shows up in the thread under your name.
Built to do real work in your help desk
- 28
- actions across 9 areas of ThriveDesk
- 5
- AI clients supported, plus any MCP client
- 60
- actions a minute, so a runaway assistant is throttled
- 1
- address for every customer, with no API key
Put your AI to work on the queue
Whether you answer tickets all day or run the team that does, ThriveDesk MCP makes the assistant you already use useful for support.
For team leads and admins
AI access, on your terms
Each person connects their own assistant, and it can never do more than they can. You decide how far it goes.
- Your own role and inbox access always come first
- Read-only, Standard, Full access presets
- Every action shows in the thread under your name
- One workspace per connection, revoke in one click
Use cases
What will you ask it first?
No syntax to learn. Say what you’d say to a teammate, and the assistant turns it into the right ThriveDesk action.
- Queue check
“What is still open in the Billing inbox from this week?”
Searches conversations, narrowed to one inbox, open status, and this week.
- Draft first
“Summarise ticket #2841 and draft a reply, but do not send it.”
Reads the thread and saves a draft on the conversation for you to approve.
- Bulk triage
“Tag every unassigned refund request with needs-finance and assign them to me.”
Tags and assigns each one. Every change shows in the thread under your name.
- Customer history
“Look up this customer and tell me what they have contacted us about before.”
Opens the contact with their recent conversations and gives you the short version.
- Team reports
“How many conversations did each agent resolve last month?”
Pulls per-agent resolution counts over the date range you named.
- Live chat
“Which live chats are waiting right now? Accept the first one.”
Lists chats in progress, then accepts one so you can reply from the same window.
Features
AI that works inside your help desk
Let the assistant pick from 28 actions
It finds the action that fits what you asked and runs it. Nothing to preload or wire up per action.
Draft first, send when you trust it
Ask for a draft and it waits on the conversation for you to approve. Grant send access later, or never.
Switch AI tools without starting over
The same address works in Claude, ChatGPT, Cursor and VS Code. Your access settings stay in ThriveDesk.
Stay in control
Pick Read-only, Standard or Full access when you connect. Revoke it from ThriveDesk at any time.
Your sign-in, not a key
You approve the connection on a ThriveDesk page with your normal sign-in. There’s no API key to paste into a chat.
Safeguards built in
60 actions a minute, no double sends on retries, and one workspace per connection.
Setup
Add ThriveDesk to your assistant
Pick your client. Whatever the menu is called, the address is always the same, and the approval happens on a ThriveDesk page in your browser.
Claude on the web, desktop, or mobile.
- 01
Open Settings, then Connectors.
- 02
Choose Add custom connector.
- 03
Name it ThriveDesk and paste the address as the URL.
https://mcp.thrivedesk.com - 04
Choose Add, then Connect. A ThriveDesk window opens for you to approve.
On Team and Enterprise plans an owner adds the connector under Organization settings first; each member then finds it under Connectors and chooses Connect.
Claude Code in your terminal.
- 01
Run this once in your terminal.
claude mcp add --transport http thrivedesk https://mcp.thrivedesk.com - 02
Type /mcp inside Claude Code, pick thrivedesk, and choose Authenticate.
- 03
Your browser opens on the ThriveDesk approval page. Pick the workspace and the access to grant.
ChatGPT on the web.
- 01
Open Settings, then Connectors.
- 02
Create a new connector and paste the address as the server URL.
https://mcp.thrivedesk.com - 03
ChatGPT sends you to ThriveDesk to approve. Choose the workspace and the access you want.
Menu wording changes between plans and versions. If you cannot find the option, a workspace admin may need to allow custom connectors first.
Cursor, per project or for every project.
- 01
Add the server to .cursor/mcp.json in your project, or ~/.cursor/mcp.json for every project.
{ "mcpServers": { "thrivedesk": { "url": "https://mcp.thrivedesk.com" } } } - 02
Open Cursor’s MCP settings and sign in when it asks.
- 03
The browser opens on the ThriveDesk approval page. Pick the workspace and the access to grant.
VS Code, per workspace.
- 01
Add the server to .vscode/mcp.json.
{ "servers": { "thrivedesk": { "type": "http", "url": "https://mcp.thrivedesk.com" } } } - 02
Start the server from the MCP view when VS Code prompts you, and confirm you trust it.
- 03
Sign-in is handled for you. The approval page opens in your browser.
Resources
Learn more about ThriveDesk MCP
Guide Set up and troubleshoot the MCP server Every client, every permission preset and every error message you might see.
Learn more
Comparison Every help desk with an MCP server Which support tools have a first-party server, and what each one can actually do.
Learn more
For AI assistants ThriveDesk facts for AI tools Plans, features and limits in one page, written for assistants that answer questions about us.
Learn more
FAQ
Frequently asked questions
Anything missing? The help article covers troubleshooting and every error message you might see.
What is MCP, in plain terms?
MCP (Model Context Protocol) is an open standard that lets an AI assistant use tools outside itself. ThriveDesk runs an MCP server at mcp.thrivedesk.com, so Claude, ChatGPT, Cursor, and VS Code can search, reply, tag, assign, and report on your real conversations instead of working from whatever you paste into the chat.
Which AI tools work with the ThriveDesk MCP server?
Claude on the web, desktop, and mobile, plus Claude Code; ChatGPT; VS Code; and Cursor. Any client that supports remote HTTP MCP servers uses the same address. If yours is not listed, contact support and we will tell you whether it is supported yet.
Which ThriveDesk plans include the MCP server?
Standard and Unlimited. Basic does not include it. On those plans any role can connect, and your role decides what you are able to grant.
Do I need an API key or a developer to set it up?
No. There is nothing to install, no API key to generate, and no developer setup. You need a ThriveDesk account on the Standard or Unlimited plan that belongs to at least one workspace, an AI client that supports remote MCP servers, and a browser to approve the connection with your normal sign-in.
Can it email my customers?
Only if you let it. Replying to a conversation, replying in live chat, and forwarding all send immediately. Internal notes, drafts, and new conversations never reach the customer; creating one only stages it for you to review. Until you trust how an assistant behaves, grant Read-only or ask it to draft, and the draft waits in the conversation for your approval.
Does this give an AI company access to all our customer data?
No. The connector answers specific requests your assistant makes on your behalf, within the permissions you granted, for the one workspace you chose. It does not hand over a copy of your help desk, and nothing is transferred in bulk or in the background.
Will my teammates see what the assistant does?
Yes. Replies, notes, assignments, and status changes land in the conversation thread under your name, the same as work you do in the app.
Can an administrator connect it for the whole team?
No. Each person connects their own assistant, because every action runs as that person and respects their inbox access. There is no shared or workspace-wide connection, and a connection covers one workspace; add a second connector for another.
How do I disconnect an assistant?
In ThriveDesk, open Apps and choose the Authorized OAuth Apps tab. Each connected assistant has its own card with the date you connected it. Choose Revoke and access stops immediately; the assistant loses every ThriveDesk action until someone authorises it again.
