Model Context Protocol

One URL.
Your AI runs the
whole workspace.

Cloodot ships a hosted MCP server — the open standard that lets Claude, ChatGPT, Cursor, and friends use real tools. Connect it once and your assistant can answer customers, update the CRM, publish campaigns, and build skills. Your sign-in, your permissions, every call logged.

MCP endpointhttps://developers.cloodot.com/mcp
Connect your client →Get a workspace first

Streamable HTTP · OAuth 2.1 · No API keys in config files

Connect

Pick your client.
Two minutes to connected.

Claude

Web · Desktop · Mobile

Add Cloodot as a custom connector on claude.ai or Claude Desktop.

  1. Open Settings → Connectors in claude.ai or the Claude desktop app.
  2. Click Add custom connector, name it Cloodot, and paste the server URL.
  3. Click Connect, sign in with your Cloodot account, and approve the access you want Claude to have.
  4. In any chat, open the tools menu and enable Cloodot — then just ask.

Custom connectors are available on paid Claude plans (Pro, Max, Team, Enterprise).

Server URL
https://developers.cloodot.com/mcp

Claude Code

CLI

One command in your terminal, one OAuth hop in the browser.

  1. Add the server. Append --scope user to make it available in every project instead of just the current one.
  2. Inside a session, run /mcp, pick cloodot, and choose Authenticate — your browser opens for the Cloodot sign-in.
  3. Done. Ask Claude Code to list conversations, draft FAQs, or deploy a skillset.
Terminal
claude mcp add --transport http cloodot https://developers.cloodot.com/mcp

ChatGPT

Web · Desktop

Add Cloodot as a connector with developer mode enabled.

  1. In ChatGPT, open Settings → Connectors → Advanced and enable Developer mode (required for connectors whose tools can take action, not just search).
  2. Back in Connectors, choose Create, name it Cloodot, paste the MCP server URL, and pick OAuth for authentication.
  3. Save, then complete the Cloodot sign-in when prompted.
  4. In a new chat, enable the Cloodot connector from the tools menu and ask away.

Connectors with developer mode require a paid ChatGPT plan.

MCP server URL
https://developers.cloodot.com/mcp

Cursor

IDE

Two JSON lines in mcp.json — Cursor handles the OAuth.

  1. Open Cursor Settings → MCP and add a new server — or drop the config straight into ~/.cursor/mcp.json (use .cursor/mcp.json inside a repo to share it with your team).
  2. Cursor shows the server with a Needs login action — click it and finish the Cloodot sign-in in your browser.
  3. Cloodot’s tools now show up in the agent’s toolset, ready to call.
~/.cursor/mcp.json
{
  "mcpServers": {
    "cloodot": {
      "url": "https://developers.cloodot.com/mcp"
    }
  }
}

VS Code

IDE · GitHub Copilot

Add an HTTP server from the Command Palette or a workspace file.

  1. Open the Command Palette and run MCP: Add Server…, choose HTTP, paste the URL, and name it cloodot — or commit a .vscode/mcp.json so the whole repo gets it.
  2. Start the server when prompted and allow the authentication flow — VS Code walks the OAuth sign-in with your Cloodot account.
  3. Open Copilot in agent mode, check the tools picker, and Cloodot is on the list.
.vscode/mcp.json
{
  "servers": {
    "cloodot": {
      "type": "http",
      "url": "https://developers.cloodot.com/mcp"
    }
  }
}

Windsurf

IDE

Point Cascade at the server URL and refresh.

  1. Open Windsurf Settings → Cascade → MCP servers and add a custom server — or edit ~/.codeium/windsurf/mcp_config.json directly.
  2. Hit Refresh in the MCP panel and complete the Cloodot sign-in when the browser opens.
  3. Cascade can now read, write, and act on your workspace.
~/.codeium/windsurf/mcp_config.json
{
  "mcpServers": {
    "cloodot": {
      "serverUrl": "https://developers.cloodot.com/mcp"
    }
  }
}

Gemini CLI

CLI

One command — or one settings.json entry.

  1. Add the server from your terminal, or put it in ~/.gemini/settings.json yourself.
  2. Launch gemini and run /mcp to check the connection — complete the OAuth sign-in if it’s pending.
  3. Gemini now carries your Cloodot toolbox into every session.
Terminal
gemini mcp add --transport http cloodot https://developers.cloodot.com/mcp
~/.gemini/settings.json
{
  "mcpServers": {
    "cloodot": {
      "httpUrl": "https://developers.cloodot.com/mcp"
    }
  }
}

Everything else

Zed · LM Studio · Perplexity · any MCP client

Any client that supports remote MCP servers takes the URL directly. For clients that only launch local (stdio) servers, the open-source mcp-remote bridge does the translation.

  1. Add Cloodot wherever your client keeps MCP servers, using the bridge config shown here.
  2. On first run the bridge opens your browser for the Cloodot OAuth sign-in, then keeps tokens refreshed.
  3. That’s it — the client sees the same 77 tools as everyone above.
mcp config (any client)
{
  "mcpServers": {
    "cloodot": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://developers.cloodot.com/mcp"]
    }
  }
}
What it can do

77 tools.
Real work, not read‑only.

The same operations your team performs in the dashboard, exposed as typed, schema-validated tools. A few from each family:

Conversations

Read threads, reply inside the 24-hour window, assign to a teammate or back to the AI, and close the loop.

  • list_conversations
  • send_conversation_messageacts
  • assign_conversation_to_ai
  • resolve_conversation

Contacts & CRM

Search customers, create and update records, and move contacts through your pipeline stages.

  • list_contacts
  • create_contactacts
  • update_contact
  • list_contact_stages

Campaigns & templates

List approved WhatsApp templates, message one customer, or publish a broadcast to thousands.

  • list_message_templates
  • send_template_oneoffacts
  • send_template_campaignacts

Knowledge & FAQs

Grow the knowledge base your AI persona answers from — FAQs, documents, and watched web pages.

  • create_faqacts
  • create_knowledgebase_weblink
  • search_documentation

Collections & catalog

Curate the product cards customers see as carousels in chat — items, media, buttons, prices.

  • list_collections
  • create_collection_itemacts
  • update_collection_item

Skills & personas

Build and deploy custom skills, install skillsets from the marketplace, and version your AI persona.

  • generate_skillsetacts
  • create_skillset_version
  • install_skillset
  • save_ai_persona_version

Plus channels, labels, canned responses, custom fields, org settings, and live documentation search — the full catalog is in the docs.

Access & safety

Your account.
Your rules.

OAuth 2.1, scoped

Clients sign in through the standard OAuth flow and get only the scopes you approve — read:conversations, write:contacts, and so on. No workspace-wide API key ever sits in a config file on disk.

Confirmation before damage

Destructive tools — deleting a contact, removing a skillset — stop and ask first. The confirmation goes to you, the human, via the client’s own approval prompt, not to the model.

Logged & revocable

Every call appears in Settings → Developers → Logs with the same per-organization rate limits as the REST API. Revoke a client’s access at any time and it’s out — mid-conversation if need be.

Good questions,
straight answers.

What is MCP, in one paragraph?
The Model Context Protocol is an open standard that lets AI assistants use real tools — the same way USB-C lets any device use any charger. A client (Claude, ChatGPT, Cursor…) connects to a server (Cloodot), discovers the tools it offers, and calls them mid-conversation. You keep one workspace; every AI you use can work in it.
Which AI apps can connect to Cloodot?
Any client that speaks MCP. Claude (web, desktop, and mobile), Claude Code, ChatGPT, Cursor, VS Code with GitHub Copilot, Windsurf, and Gemini CLI connect natively over streamable HTTP. Clients that only support local stdio servers — Zed, LM Studio, and others — connect through the mcp-remote bridge with one config entry.
Is it safe to let an AI touch my workspace?
You stay in charge at three layers. Sign-in is OAuth 2.1 — you approve exactly which scopes a client gets, and no API key is ever pasted into a config file. Destructive tools (like deleting a contact) require an explicit confirmation before they run. And every call is logged in Settings → Developers → Logs, where access can be revoked at any time.
Do I need an API key?
No. The MCP server authenticates with OAuth: your client opens a browser window, you sign in to Cloodot, and scoped tokens are issued behind the scenes. API keys exist too, but they are for the REST API at developers.cloodot.com — MCP never needs one.
Does MCP access cost extra?
No. Your AI works under the same workspace, plan limits, and rate limits as your team in the dashboard. There is even a get_plan_limits tool, so your assistant can check remaining headroom before it creates anything.
What exactly is the server URL?
https://developers.cloodot.com/mcp — streamable HTTP with OAuth 2.1. Agents can also discover it automatically via the server card at cloodot.com/.well-known/mcp/server-card.json.
Get started

Give your AI
a workspace.

Start a free Cloodot trial, paste one URL into the assistant you already use, and watch it do real work — with your permissions and a full audit trail.

Start free trial →Read the developer docs
No credit card requiredConnect in two minutesRevoke anytime