Connect CluGTM to Claude (MCP)

The CluGTM MCP (Model Context Protocol) connector lets Claude use CluGTM as a set of tools. You describe what you want in plain language, for example “find heads of RevOps at Series B fintechs in New York and get their work emails”, and Claude runs the search, enriches the shortlist and saves the results to a lead list, all from the chat.

It works the same in claude.ai, Claude Desktop, Claude Code and Cowork. The MCP tab in CluGTM is labelled Beta.

Who can use it

MCP is separate from the public REST API. API keys are a Pro and Enterprise feature (and AppSumo Tiers 5 to 9), but you don’t need an API key to use the MCP connector. See What’s included by plan for the full matrix.

Connect in one click

1. In CluGTM, open Integrations in the sidebar and choose the MCP tab (or go to app.clugtm.com/integrations/claude ). The panel is titled Connect to Claude.
2. Under CluGTM MCP URL, click Copy URL. The line under the URL reads “Connects Claude to (your workspace name). Add one connector per workspace.”
3. Open Claude: go to Settings → Connectors. This works the same in claude.ai and the Claude desktop app, and the connector then appears in the mobile app and Cowork too.
4. Add a custom connector: click Add custom connector, name it CluGTM and paste the CluGTM MCP URL. Leave Authentication on Sign in now and OAuth client on Use Claude’s published identity, then click Add.
5. Sign in with your CluGTM account: click Connect. Claude opens a CluGTM page in your browser. Sign in if asked, then review what Claude is asking to access (your identity, basic profile, email address, public metadata and organization membership) and click Allow.

That’s it. Claude can now call CluGTM tools in any conversation, and you can turn the connector on or off per conversation from Claude’s tools menu. There is no OAuth Client ID or client secret to paste, and no API key to create.

The URL looks like https://mcp.cleanlist.ai/v1/mcp?org=org_.... Always copy it from the MCP tab rather than typing it, because the org= part identifies your workspace.

ChatGPT and other apps

The same URL works in ChatGPT and other apps that support custom MCP connectors, such as Cursor and Codex. Add it as a connector in that app, then sign in with your CluGTM account and allow access, as in step 5. Everything on this page about workspaces, tools, costs and approvals applies the same way.

One connector per workspace

Each connector acts in the workspace named in its URL, no matter which workspace you have open in the CluGTM app.

• Several workspaces: switch to the other workspace in CluGTM, open Integrations → MCP again and click Copy URL. The URL changes with the workspace. Add it to Claude as a second connector.
• Credits and data: everything Claude does is charged to, and saved in, the connector’s workspace. Claude acts as you, with your access in that workspace.
• Older connectors: if your connector URL has no ?org= part, it was added before connectors were tied to a workspace. Remove it in Claude and add it again with the URL from the MCP tab.

Tool groups

Tools come in groups. A connector gets the default groups unless its URL asks for others. To get every group, add tools=all to the URL (for example ...?org=org_...&tools=all). To pick groups, list them: tools=agents,crm. Workspace basics are always included.

A tool your plan doesn’t include is left out even when its group is in the URL. Workspaces on an AppSumo deal without a paid CluGTM plan get only the core set of 14 tools, whatever the URL asks for (see Troubleshooting).

The connector returns primary work emails only, in every group. Personal emails aren’t available through it; use Find Personal Email in the app.

Each tool tells Claude whether it only reads, changes data, or deletes it. Claude can run read-only tools, such as searches, without asking each time, and always asks before a tool that deletes data or writes to an outside system such as your CRM.

More tools in the chat

Claude can also load Clu’s own tools during a chat with the load_tools tool, one group at a time: lists, skills, agents (your GTM agents), plays, sequences and crm. These are the actions of the API, and they’re separate from the tools= groups in the URL. A group lists only the actions your plan, connections and sign-in allow; when some need a scope your sign-in doesn’t hold, the group says which. A loaded group stays for that chat; if a tool from an earlier turn is missing, Claude loads its group again.

An action that spends credits, sends, deletes, writes to your CRM or turns an agent on doesn’t run when Claude calls it. It returns a card with what it will do and what it costs. Claude shows you the card and ends its turn, and runs it with the confirm tool only after you say yes. Once the lists group is loaded, deleting a list or removing leads goes through these cards.

Deletes, sends, CRM writes and anything that turns an agent on, runs it now or changes its limits also need your own approval in the app’s approval prompt, not just a yes in the chat. claude.ai, Claude Desktop, ChatGPT, Cursor and Codex show it. Claude Code doesn’t, so these actions don’t run there: Claude tells you to confirm them in CluGTM instead. If you decline or close the prompt, nothing runs.

What Claude can do, and what it costs

The tools draw on the same data, credit balance and lists as the CluGTM app. Claude only sees the tools your plan, workspace and tool groups allow, so the exact list varies.

Contacts Claude enriches without a target list are saved to a protected list named MCP Results in your Lead Lists.

The connector covers search, lists, enrichment, skills, sync and export, and Claude can load more tools for your agents, plays, sequences and CRM. Clu itself stays in the CluGTM app.

Quotes and approval

Claude never starts paid bulk work on its own:

1. Before enriching a list or a set of search results, running a skill, or syncing leads, Claude asks CluGTM for a quote (the estimate_cost tool).
2. Claude shows you the credit cost and waits for your approval.
3. The action runs with that quote. A quote lasts 5 minutes and works once. If it expires, or the selection grows by more than 10%, Claude gets a new quote first.

Syncing to HubSpot, Salesforce or Lemlist always needs your explicit approval.

Rate limits

Each workspace can make 60 requests a minute through the connector and the API combined (30 a minute for workspaces on an AppSumo deal without a paid plan), and up to 60 people searches a minute. If Claude reports a rate limit, wait a minute and ask again.

Example prompts

Find people and save them to a list

Find VPs of Sales at US-based SaaS companies with 50-200 employees, and save the first 25 to a new list called “Q4 SaaS VPs”.

If something material is missing (for example, which regions count), Claude asks one short follow-up. It then runs the search, shows the matches as a table, creates the list and saves the people you picked.

Enrich a list

Enrich my “Q4 SaaS VPs” list with work emails.

Claude gets a quote for that list and shows you the credit cost and your balance. After you approve, it starts the enrichment, reports progress, and summarizes how many contacts got emails.

Check the workspace

Which lead lists do I have, how many leads are in each, and how many credits do I have left?

Claude lists your lead lists with their lead counts and shows your credit balance. Nothing changes and no credits are spent.

Qualify leads (needs tools=agents)

Score the leads in “Q4 SaaS VPs” against my ideal customer profile and show me the top 10.

Claude first tells you what your ideal customer profile says, since every score is measured against it. It then quotes the per-lead cost, waits for your approval, runs the Qualification skill, and shows the best-fitting leads with a score, a verdict and the reasons.

Disconnecting

To disconnect, remove the CluGTM connector in Claude under Settings → Connectors. Removing it stops Claude from calling CluGTM; your lists and enrichment results stay in CluGTM.

Troubleshooting

Still stuck? Email support@cleanlist.ai with the workspace name and what you asked Claude to do.

For developers

The connector is hosted by CluGTM at mcp.cleanlist.ai and calls the same backend as the v2 REST API. Sign-in uses OAuth with your CluGTM account, so no API key is involved.

Working in Claude Code, Cursor or Codex, or with CSV files and scripts? The CluGTM CLI (npx @cleanlist-ai/cli) runs the same search, enrichment, list and export work from the terminal. Rows move between files and CluGTM without passing through the conversation. To call CluGTM from your own code, use the REST API with an API key: start with the v2 quickstart.

Privacy and data

• The connector reaches your CluGTM workspace through your own account, after you approve access. It can only see and change what your account can see and change in the app.
• CluGTM receives the inputs Claude sends to its tools (for example, search criteria or a list name), not your whole conversation.
• Your workspace data is handled as described in the CluGTM privacy policy . Questions: privacy@cleanlist.ai.

Related

• Integrations overview
• What’s included by plan
• Pricing and credits
• Skills
• Find People
• Clu, the in-app assistant
• CluGTM CLI for coding agents
• API reference (v2)
• Tutorials