Connect Salesforce to Claude. Your Salesforce stays current and nobody types into it.
Claude keeps your Salesforce current: stage, next step, close date, contacts, notes, written from the calls and emails around each deal. Reps stop typing into it. The forecast reflects what happened on the deals, not what someone remembered to log.
Tell us how your team works today. We build a demo on your workflow and send you a recorded walkthrough within 3 business days.
Sample data, same dashboards we run on.
You run Salesforce. This is built for you.
Yes. This is the build for teams that already run Salesforce.
Your Salesforce stays your system of record. Your licenses, your admin, your reports and your dashboards stay as they are. Claude works with the org you have. Nothing here asks you to change what Salesforce is.
What changes is who does the typing. Today the record is only as current as the last time a rep updated it. With Claude connected the right way, the record is updated from the call, the email thread and the meeting, and every change is logged.
This page covers how to connect Claude to Salesforce, four ways, and which one holds up for a team.
For sales leaders and for RevOps
For sales leaders
A forecast you can trust, because each number points to the call or email behind it.
Pipeline review without the cleanup. The meeting starts with the deals, not with fixing fields.
For RevOps
Records stay complete without chasing reps. Reps are not typing, and the fields still get filled.
One data model you can see and change. Your Salesforce admin keeps control of objects, fields and reports.
Time back for analysis instead of cleanup.
You own the data model and the dashboards.
What you get when Claude is connected to Salesforce
Most Salesforce orgs hold the answers leadership asks for every week. What is in the pipeline? What is the real forecast? What happened on that account last month? Reps still spend time pulling those answers from list views and reports, and updating fields by hand.
Four things change once Claude is connected well:
Pipeline answers in plain language.
“Which deals haven’t moved in three weeks?” is answered in a sentence, with the deals listed.
A forecast you can trust.
Claude reads the same records your team does, plus the calls, emails and notes around them. The number reflects what is happening on each deal, not only what is typed in a field.
Reps stop typing.
The CRM fills in from calls, emails and meetings, and your RevOps team still owns the data model. Fields are updated without anyone being asked to remember.
Meeting prep with full account context.
Before a call, Claude assembles the account history, open opportunities and recent activity in one briefing.
Four ways to connect Claude to Salesforce
There are four ways to connect Claude to Salesforce today. The first three decide how Claude reaches Salesforce. The fourth decides what Claude reads once it is connected. It is the one we build for teams that use this every day.
Option 1
A direct MCP connector
MCP (Model Context Protocol, see our MCP glossary entry) is the open standard Anthropic published for connecting Claude to tools and data. A connector that exposes Salesforce objects as tools is the fastest way to get Claude talking to Salesforce. Salesforce also documents hosted MCP servers for connecting agents to Salesforce (see Salesforce’s developer documentation, “Hosted MCP Servers”).
When it is right
Early evaluation, one power user, read access to a sandbox org.
Limit
A generic or community connector has no shared permission boundary, no central log of what was read or changed, and no rate-limit handling. It works for one person and has no audit trail when a team starts using it.
Option 2
Direct API integrations and AI-written glue code
A script that calls the Salesforce REST API and passes the result to Claude.
When it is right
A prototype, a personal automation, a proof of concept.
Limit
Glue code has no retry logic and no versioning, and usually one person understands it. It breaks when a field is renamed, an API limit is hit or its author leaves. It rarely passes a security review, because a script with standing access to CRM data is not an answer a security team accepts.
Option 3
Claude through MuleSoft to Salesforce
Claude calls MuleSoft-served MCP tools. MuleSoft handles authentication, logs every call, transforms data, and applies retry logic, rate limits and versioning on the way to Salesforce. MuleSoft’s Agent Fabric is Salesforce’s product version of this approach.
When it is right
Any connection a team relies on, any connection that touches production data, any connection that has to pass a security review.
Limit
Claude can read and update Salesforce records, and that is all it can see. Whatever was said on the call and never typed into a field is invisible to it.
Option 4 · The build we recommend
Claude reads Salesforce and the conversations around it
We start with the controls from Option 3 and add a second source for Claude to read: the calls, email threads, meeting history and decisions behind your deals. Claude reads both together and keeps your Salesforce current.
When it is right
Teams that want Claude doing real work (pipeline analysis, forecasting, meeting prep, keeping fields current) and not only looking up records.
Limit
It takes a short build with you to decide which Salesforce objects and which conversations go in, and which fields Claude may write on its own and which wait for a person to approve.
Why not call Salesforce live every time? Salesforce’s API returns one record at a time and is subject to API limits. It holds what reps typed. It does not hold what was said. A question like “which deals are stalling, and why?” needs the pipeline and the conversations around it, read together.
Salesforce loses nothing in this pattern. It stays the system of record. Every update is logged and attributable. The org also gets more complete, because updates come from the call and the email, not from memory at the end of the week.
Everything runs in your accounts. We run no servers for you.
When one agent becomes several, MuleSoft Agent Fabric is how teams manage them. Governance is not optional here. An auditor who asks “who changed this opportunity stage and why” needs a logged answer that is not “an AI decided.” A security team needs a logged, rate-limited, versioned call path and not a script.
We do not recommend a pattern we have not used. Green Irony runs its own company on Claude, and the pattern on this page is the one we use daily.
Our fixed-bid quoting used to take weeks of back-and-forth. It now takes minutes. This is our own experience, not a promise for your team. We use the same approach to screen candidates and to flag customers at risk of leaving. In each case Claude works from our records and the conversations around them, with every action logged.
Claude is only as accurate as the org it reads. Duplicate accounts, stale stages, contacts with no owner and fields nobody has filled in for two years produce confident answers that are wrong.
Before connecting anything, check the org. Our Salesforce Reviver health check shows what is duplicated, stale, missing and unused. Our demo also starts with how your org is set up today, based on what you tell us in the form.
Get a demo built on your workflow
Tell us how you use Salesforce today and which workflow to show you. We build the demo around that workflow.
Then, if you want, we walk through it live with you. Already have Claude connected to Salesforce and want a second opinion before a security review? Say so on the form. We do that too.
Occasional email about running a company on Claude, with the 90-second demo and short, plain notes. No weekly filler.
Unsubscribe any time. We never share your address.
Frequently asked questions
How do I connect Salesforce to Claude?
There are four ways. A direct MCP connector is the fastest and works for one person exploring. A script that calls the Salesforce API works for a prototype. For anything a team relies on, connect through MuleSoft, which handles authentication, logging, rate limits and retries. For daily use, add your calls, emails and decisions as a second source, so Claude reads both. Salesforce stays the system of record in every case.
What is the Salesforce MCP connector?
MCP, the Model Context Protocol, is the open standard Anthropic published for connecting Claude to external tools and data. A Salesforce MCP connector exposes Salesforce objects and actions as tools Claude can call. Salesforce also publishes MCP servers and APIs for agents. A connector on its own is useful for evaluation. A team needs logging, permission boundaries and rate-limit handling around it, and those come from how it is connected.
Is it safe to connect Claude to CRM data?
It is safe when the connection logs every call, enforces authentication and applies rate limits and retry logic, the same controls you expect from any integration that touches customer data. It is not safe when Claude has a direct line into the CRM with no audit trail, because that setup fails a security review and leaves no record of what was read or changed.
Should Claude read Salesforce directly, or also read a record of my calls and emails?
For quick lookups, a direct connection is fine. For pipeline analysis, forecasting and meeting prep, Claude needs the conversations as well as the records. Salesforce's API returns one record at a time and holds what reps typed. Adding your calls, emails and decisions as a second source gives Claude both. Every write to Salesforce is logged.
Can Claude update Salesforce records?
Yes. Claude can create and update Salesforce records, with every write logged and attributable. Claude works with the same permissions as the person using it. Your team decides which fields it can update. Unlogged, ad hoc API calls leave no trace of what changed or why, so we do not use them.
Do I need MuleSoft to connect Claude to Salesforce?
No. A direct MCP connector or a script gets Claude talking to Salesforce. You need MuleSoft, or an equivalent integration platform, when the connection has to pass a security review, serve more than one user, or answer "who accessed what, and when."
What can Claude do once it is connected to Salesforce?
Claude can answer pipeline questions in plain language, produce a forecast based on real account activity, fill in CRM records from calls and emails so reps stop doing manual data entry, and assemble full account context before a meeting. Salesforce stays the system of record. Claude makes what is in it usable.
How much does it cost to connect Claude to Salesforce?
It depends on which of the four ways you choose and how much of your org needs cleanup first, so we scope it after the demo and do not list a number here. We do not publish a number because the scope varies too much.
We run Salesforce. Is this for us?
Yes. This is the build for teams that already run Salesforce. Claude plus Salesforce keeps your Salesforce current without anyone typing into it. Salesforce stays your system of record, and your licenses, admin and reports stay as they are.
Where does this run, and who hosts it?
In your own accounts. Everything runs in accounts you own, and we run no servers for you. We build it and, if you choose, operate it for you.
What do you need from me to start?
A short request form at /demo/. Tell us which CRM you use, which workflow to show you (for example forecast call prep or deal updates after calls), and how your team does it today. You can also upload a redacted deal, a process doc or a screenshot. We build the demo around it.
