Overview
Your CRM tells you a deal moved to stage 4. Weflow tells you why, in the rep's own words, scored against your own playbook. Now that layer is available to Claude, ChatGPT, and anything else your team builds on.
This MCP should be used in combination with a MCP connector by Salesforce to make use of the full extent of your SFDC data in an AI Assistant, Warehouse or otherwise.
Every AI tool can read your CRM by now. That is not the interesting part.
The interesting part is what your CRM never captured. What the champion actually said about budget. Whether the rep asked the qualifying question at all. What got committed on the forecast call and what quietly changed by Friday.
Weflow now ships an MCP server for exactly that layer: call intelligence, forecast calls, and your AI playbooks. Any MCP client can read it. Ask Claude to prep a deal review and it works from the transcripts and your methodology scores, not a stage field.
Setting up the MCP in e.g. Claude
Step 1: Copy your MCP server domain from Weflow
In Weflow, go to Admin Console > Weflow AI > MCP Server and copy the server domain shown there. This URL is unique to your workspace. You need admin rights in Weflow to see it.
Step 2: Add it as a custom connector in Claude
On a Pro or Max plan, open Settings > Connectors in Claude, click the + button, and choose Add custom connector. Give it a name (Weflow works fine), paste the domain you copied, and click Add. Claude will then ask you to connect, which sends you through a sign in to authorize access.
On a Team or Enterprise plan, an Owner has to add the connector for the organization first, under Organization Settings > Connectors > Add > Custom, choosing Web as the type if prompted. After that, each team member connects individually. See the next section for what that means.
Every user authenticates as themselves
Adding the connector and connecting to it are two different things, and the distinction matters.
When an Owner adds Weflow under Organization Settings, that is registration only. It tells Claude that a server exists at your workspace URL. It grants no access to any data, for anyone, including the Owner. Nothing is readable yet.
Each user then connects individually. They open Settings > Connectors, find Weflow in the list their Owner added, and click Connect. Claude sends them to Weflow, they sign in with their own credentials, and they see a consent screen showing exactly what Claude is asking for. On approval, Weflow issues an access token bound to that one user and returns it to Claude.
From then on, every request Claude makes carries that individual's token. Weflow resolves permissions exactly as it does when the person is clicking through the app. An AE who connects their own account cannot pull the VP's forecast calls through Claude, because they cannot pull them in Weflow either. There is no shared service account and no admin token that sees everything.
Revocation works from either end. A user can disconnect the connector in Claude, or an admin can revoke that user's token in Admin Console > Weflow AI > MCP Server. Either action cuts off that person's access immediately and leaves everyone else connected.
Step 3: Use it
Back in any conversation, click the + button to make sure the Weflow connector is enabled, then just ask. A few that work well on day one:
What did the champion say about budget on the last three ACME Inc. calls?
Score this opportunity against our playbook and tell me what is missing.
Summarize what changed between last week's forecast call and this one.