Claude
Anthropic's Claude wired into Salesforce: summarize cases, draft replies, extract data, and talk to your org through MCP. We run three Claude-powered tools in production on this site.
Claude on Salesforce
Claude connected to Salesforce turns the CRM your team already lives in into an AI workbench: cases summarized on arrival, reply drafts written from full record context, unstructured text classified onto the right fields, and an org you can query in plain English. This page covers the four ways to wire the two together and when each fits.
The four patterns: the Anthropic Messages API called from Apex or Flow for automation inside the org; Claude inside Agentforce through Amazon Bedrock for teams that want frontier AI within the Salesforce trust boundary; the Model Context Protocol (MCP), which lets Claude Desktop and Claude Code read and act on your org from a conversation; and Claude Code with our 141 free Salesforce skills for development work.
For the API pattern we build it the Salesforce-native way: an External plus Named Credential so the API key never lives in code, prompts assembled from record context and stored in Custom Metadata so admins can tune them without a deploy, async callouts that respect governor limits, and every result written back to the record with a full audit trail.
We do not just recommend this stack, we run it: Marc the site chatbot, the CRM Advisor, and the Salesforce Formula Generator on cloudsheer.com are all Claude-powered and in production. Every Claude build is delivered by a senior architect at a fixed price, sandbox-tested, and backed by 30 days of hypercare.
Quick facts
- Platform
- Anthropic Claude
- Type
- AI
- Direction
- Outbound API, two-way via MCP
- Objects
- Any record
- Best for
- Summarization, drafting, org queries
Why connect Claude to Salesforce
What your team actually gets out of the integration.
AI where the work already happens
Summaries, drafts, and classifications land on the Salesforce record, so nobody pastes CRM data into a chat window again.
Four patterns, one partner
API automation, Agentforce via Bedrock, MCP for conversational access, and Claude Code for development - we pick the one that fits, not the one we sell.
Proven in our own production
Three Claude-powered tools run live on this site. You are buying a pattern we operate daily, not a slide.
Built by senior architects
A production-grade build with governance, testing, and support, not a brittle one-off script.
Live in weeks, at a fixed price
A senior architect scopes and ships your Claude integration on a fixed price, usually in two to six weeks.
Owned by you
Documented, deployed to your own org, and yours to keep. No black-box middleware and no lock-in.
What a Claude integration changes
The measurable difference teams see once the data flows on its own, instead of through copy-paste and spreadsheets.
to a summarized case with a drafted reply waiting for review
copy-paste between Claude and the CRM
queries against your org through MCP
write-back with a full audit trail
Where teams use Claude on Salesforce
The scenarios this integration is built for, including the ones we have already delivered.
Summarize a case on arrival and draft the first reply
Extract structured fields from emails and documents
Query the org in plain English from Claude Desktop via MCP
Summarize a case on arrival and draft the first reply for agent review.
Extract structured fields from inbound emails and documents onto the record.
Query your org in plain English from Claude Desktop through MCP.
Build Salesforce faster with Claude Code and our 141 free Salesforce skills.
Classify and route inbound leads by intent before a human touches them.
Draft personalized outreach from full CRM context, written back for review.
Teams that get the most from Claude on Salesforce
The roles that feel the difference on day one.
Three ways to connect Claude to Salesforce
And which one we recommend for a build like this.
| Agentforce with Claude (Bedrock) | MCP for Claude Desktop / Code | Custom Apex + Messages APIWe recommend | |
|---|---|---|---|
| Best for | Agentic AI inside the Salesforce trust boundary | Analysts and admins working conversationally | AI wired into records and automation |
| Speed to value | Fast (licensed through Salesforce) | Fast | Moderate |
| Flexibility | Within Agentforce guardrails | High, human-in-the-loop | Unlimited |
| Ongoing cost | Agentforce licensing | API usage only | API usage plus build |
| Best when | You already run Agentforce | People ask ad-hoc questions | You want AI inside the workflow |
How Claude connects to Salesforce
The shape of the integration at a glance. For the full engineering walkthrough with real code, read the technical guide.
The connection at a glance
Data flows left to right
Read the full technical guide
Architecture, the step-by-step build with real Apex and DataWeave code, field mapping, security, and the pitfalls we design out. Written for developers and admins.
What we have built with Claude
Anthropic Claude connected to Salesforce for summarization, drafting, and conversational org access - the same models running three production tools on cloudsheer.com.
Marc, the site chatbot
A Claude-powered pre-sales assistant running in production on cloudsheer.com.
CRM Advisor
A Claude-backed recommendation tool weighing Salesforce, HubSpot, Zoho, and Dynamics.
Salesforce Formula Generator
A four-mode Claude workbench that generates, explains, fixes, and optimizes formulas.
Real components we ship
What to expect
A typical engagement, from the first call to a monitored go-live.
Discovery and scoping
2 to 4 daysWe map the use cases, pick the connection pattern, and agree a fixed price.
Build in a sandbox
1 to 3 weeksNamed Credentials, prompt design, the Apex service, and automation for the Claude integration.
Test and UAT
3 to 5 daysGolden-example prompt validation, edge cases, and a parallel run against human output.
Go-live and hypercare
30 daysProduction deployment, training, and monitored support with token-spend tracking.
Claude Salesforce integration FAQs
How do I connect Claude to Salesforce?
Four ways: the Anthropic Messages API called from Apex or Flow for in-org automation, Claude inside Agentforce via Amazon Bedrock, an MCP server so Claude Desktop can query and update the org conversationally, or Claude Code for development work. Most clients start with the API pattern: Named Credentials, an async Apex service, and results written back to the record. We scope the right pattern on a free call.
Can Claude run inside Agentforce?
Yes. Salesforce and Anthropic have a partnership that makes Claude models available inside Agentforce through Amazon Bedrock, keeping processing within the Salesforce trust boundary - aimed squarely at regulated industries like financial services and healthcare. Licensing runs through Salesforce; we handle the model configuration and the agent build.
What does the MCP option actually do?
The Model Context Protocol lets Claude Desktop and Claude Code connect to your org through a server exposing tools like SOQL query and scoped record updates. Your team asks questions in plain English - pipeline by stage, accounts with open cases - and Claude runs the queries with the permissions of a dedicated integration user, so it sees only what it is granted.
Claude or the models Agentforce ships with?
Not either-or. Agentforce's own models handle CRM-native tasks well; Claude excels at long-context reasoning, nuanced drafting, and complex extraction. We implement both and route each use case to whichever does it better - that is the advantage of a partner that is not selling you one model.
Where is the Anthropic API key stored?
In Salesforce External and Named Credentials guarded by a permission set. Nothing is hard-coded, so the key is never exposed in code, metadata, or logs.
Will our Salesforce data be used to train Claude?
No - under Anthropic's commercial API terms, your inputs and outputs are not used for model training by default. For sensitive fields we add a redaction step before anything leaves Salesforce, and the Bedrock pattern keeps processing inside the Salesforce trust boundary entirely.
How long does a Claude Salesforce integration take?
A first use case (case summarization or reply drafting) typically ships in two to four weeks; MCP setups are often faster. We scope it up front and quote a fixed price.
What can Claude actually do once connected?
Summarize cases on arrival, draft replies from full record context, extract structured data from emails and documents, classify and route leads, answer plain-English questions about your org via MCP, and accelerate development through Claude Code. The three Claude tools running on this site are live examples.
Related integrations
All integrationsNeed a Claude integration done right?
Tell us your setup and where the data breaks. In a free 30-minute call we will map the Claude flow and hand you a clear, fixed-price plan.