Connect Claude to your CRM (MCP)

MCP lets you connect your own Claude directly to your KudosCRM workspace, so you can ask about your pipeline, search records, or draft content without leaving the conversation.

It is built on the API token you already have, and it deliberately cannot do more than you can.

Who this is for

Anyone comfortable editing a configuration file, who already uses Claude.

Before you start

• An API token from your KudosCRM settings.
• Claude, configured to accept an MCP server.
• Somewhere secure to keep the token — it goes in a config file.

Writes land as drafts, on purpose
The write tools deliberately create things in a non-live state: a price book comes in inactive, a product as a draft, a workflow disabled, and a help article as a draft. That is a fixed behaviour, not a setting — so an agent can prepare work but cannot switch anything on for your customers. There is no delete tool and no bulk tool either.

Connect it

1. Mint an API token
   Create one in settings. MCP uses the same token type as the REST API rather than a new kind of credential.
2. Add the MCP server to your Claude config
   Point it at your workspace's MCP endpoint with the token as a bearer credential. It is an HTTP server, so declare it as one.
3. Confirm which workspace you are on
   Ask it who it is connected as. If you have more than one workspace, this is the check that stops you writing to the wrong one.
4. Start with read tools
   Search contacts and deals, summarise the pipeline, list products and price books. These answer questions without changing anything.
5. Use write tools knowing they land as drafts
   Creating a product, price book, workflow or help article produces something inactive that a person then reviews and publishes.
6. Revoke when you are done
   If the connection was for a one-off task, revoke the token afterwards rather than leaving it live.

What you get

• Plain-language access to your CRM from your own Claude.
• Exactly your permissions — the agent cannot exceed what your account can do.
• Writes that land as drafts for a person to review, never live.
• An audit record of every call the agent made, including the ones that were denied.

Frequently asked questions

Can the agent see data I cannot?

No. It inherits your own permissions and is checked on every call, so it is bounded by exactly what your account can already reach.

Can it publish something to my customers?

No. Writes are forced into a non-live state — inactive, draft or disabled — and there is no publish tool. A person has to make anything live.

Can it delete records?

There is no delete tool and no bulk tool. The write surface is deliberately narrow.

Is there a record of what it did?

Yes. Every call is audited, including calls that were denied, which is usually the more interesting half.