Claudeforce
by Salesforce (with Anthropic)
Salesforce CRM inside Claude — 37 prebuilt sales skills with permissions inherited from your org
Claudeforce is the Salesforce–Anthropic partnership that brings CRM data and governed actions into Claude. Its first product, Salesforce in Claude, is a plugin with 37 prebuilt sales skills letting sellers run meeting prep, deal-health checks and pipeline reviews and write updates back to Salesforce without leaving Claude, while Claude simultaneously becomes the default reasoning model inside Agentforce.
Claudeforce is the joint Salesforce and Anthropic offering announced on 26 August 2026 that puts Salesforce CRM data and workflows directly inside Claude, and makes Claude the default reasoning model inside Salesforce's own agent stack. Its first shipping component, Salesforce in Claude, is a Claude plugin carrying 37 prebuilt sales skills — meeting preparation, deal-health review, pipeline review and similar — that let a seller query live revenue context, assemble tailored dashboards and push governed updates back into the CRM without opening the Salesforce application. The plugin runs on AIforce, Salesforce's enterprise harness that exposes business data and workflows to any agent through hosted MCP servers, APIs and CLI tools; Claude calls the Salesforce MCP server, which inherits the signed-in user's existing record permissions, so a user cannot reach data they could not already see, and administrators configure access once rather than per user. The prebuilt skills matter because raw MCP access gives a model only general-purpose SOQL and CRUD tools that it must improvise a workflow around every time, whereas a skill encodes the intended sequence as human-readable instructions. Running the other direction, Claude becomes the default model in Agentforce, powering Agentforce Vibes and Agentforce Coworker, delivered through Amazon Bedrock inside the Salesforce trust boundary so data and inference stay within enterprise controls. Availability is staged: select pilot customers at announcement, open beta in September 2026, and additional service, marketing and commerce skills from late 2026. Commercially it reflects a shift toward consumption pricing, with Salesforce charging headless API consumption that varies by licence edition while Claude inference is contracted separately with Anthropic. Salesforce cites its own internal usage as evidence, reporting that 83% of its workforce uses a Claude-powered Slackbot.
A Salesforce enterprise whose sellers and revenue-operations staff already live in Claude all day, and where CRM hygiene suffers because nobody wants to open the Salesforce UI to log or check anything.
Sellers run meeting prep, deal-health and pipeline reviews and write governed updates back to the CRM from inside Claude, under the Salesforce permission model the organisation already audits.
At a Glance
- Category
- AI Agents & Orchestration
- Pricing
- Usage-based, Contact for pricing
- Target Market
- CIOs, CTOs, Sales Leaders, Revenue Operations, Salesforce Administrators
- Deployment
- Cloud-only, API-based
- Founded
- 1999
- Headquarters
- San Francisco, United States
- Team Size
- 500+
- Customers
- Select pilot customers as of August 2026; open beta scheduled for September 2026
Key Features
- ✓37 prebuilt sales skills
Covers meeting preparation, deal health review and pipeline review out of the box rather than requiring prompt engineering.
- ✓Salesforce hosted MCP servers
Expose live CRM records to Claude with permissions inherited from the signed-in user, so record-level access rules still apply.
- ✓AIforce enterprise harness
Surfaces Salesforce data and workflows to any agent through MCP servers, APIs and CLI tools without custom integration work.
- ✓One-time administrator setup
Replaces per-user MCP configuration, removing the account-by-account integration friction that stalls comparable rollouts.
- ✓Claude as default Agentforce model
Claude becomes the default reasoning model in Agentforce, powering Agentforce Vibes and Agentforce Coworker.
- ✓Amazon Bedrock inside the trust boundary
Claude inference is delivered via Bedrock within the Salesforce trust boundary so data stays under existing enterprise controls.
- ✓Governed write-back to CRM
Pushes pipeline and record updates into Salesforce with existing business rules and permissions enforced on every write.
Capabilities
Use Cases
- •Automated meeting preparation
Have Claude assemble account history, open opportunities and recent activity before a customer meeting without manual lookup.
- •Deal health review
Run a health check across the pipeline and surface at-risk opportunities without opening the Salesforce console.
- •Conversational CRM updates
Update opportunity stages, next steps and close dates from a Claude conversation with permissions and business rules enforced.
- •Ad-hoc pipeline dashboards
Let a sales manager build a pipeline view from live CRM data with a single natural-language request instead of a report builder.
- •Unified Slack and CRM context
Reason over Slack conversations and Salesforce records together in one place rather than switching between tools.
Ideal For
Best For
- ✓Salesforce enterprises whose sellers already work day to day inside Claude
- ✓Revenue operations teams wanting CRM hygiene without forcing sellers back into the Salesforce UI
- ✓Existing Agentforce customers who want Claude as the default reasoning model
- ✓Organisations standardising on MCP as the access layer for enterprise data
- ✓Large sales organisations where per-user integration setup has blocked previous AI rollouts
Not Ideal For
- ✗Anyone who needs this in production today — it is limited to select pilot customers, with open beta only scheduled for September 2026 and no announced general availability date
- ✗Non-Salesforce CRM shops, since the entire value rests on Salesforce data, metadata and years of encoded workflows
- ✗Buyers requiring a single predictable line item: Salesforce consumption and Anthropic inference are contracted separately and no combined price exists
- ✗Teams needing service, marketing or commerce coverage now — only sales skills ship at launch, with the rest deferred to late 2026
- ✗Organisations standardised on a different model provider, given Claude is the default reasoning model here
Integrations
Deployment
Market & Ratings
Select pilot customers as of August 2026; open beta scheduled for September 2026
Market Analysis
Pros
- ✓Permissions inherit from Salesforce, so the security model is the one the enterprise already audits rather than a new surface to review
- ✓Prebuilt skills encode the workflow, which is materially more reliable than letting a model improvise SOQL and CRUD against raw MCP every time
- ✓One-time administrator setup avoids the per-user MCP configuration that has stalled comparable enterprise AI rollouts
- ✓Delivered through Amazon Bedrock inside the Salesforce trust boundary, which addresses the data-egress objection directly rather than deflecting it
Cons
- ✗Not generally available: pilot-only at announcement, open beta targeted for September 2026, and no GA date has been given
- ✗Pricing is genuinely unresolved — Salesforce charges headless API consumption varying by licence edition while Claude inference is a separate Anthropic contract, and a Salesforce executive admitted you cannot buy it on one piece of paper
- ✗Salesforce community coverage flags unanswered administrator questions: whether the 37 skills are customisable, how business rules are actually configured and enforced, and what audit logging looks like
- ✗Only sales is covered at launch; service, marketing and commerce skills are deferred to late 2026, so most of the CRM is out of scope for now
- ✗Analysts note a strategic risk that moving the interface into Claude shifts pricing power toward the model provider and cannibalises Salesforce's seat-based revenue faster than consumption replaces it
Pricing
Pilot / open beta
Contact for pricing
- ✓Salesforce in Claude plugin with 37 sales skills
- ✓Salesforce hosted MCP server access
- ✓Headless API consumption metered by licence edition
- ✓Claude inference contracted separately with Anthropic
No combined price exists. Salesforce meters this as headless API consumption whose allowance varies by licence edition, while Claude inference is contracted separately with Anthropic — a Salesforce executive conceded you cannot buy it on one piece of paper at the moment. In practice that means an existing Salesforce licence plus a separate Anthropic commitment, with consumption forecasting unresolved and no published rate card on either side as of August 2026.
Security & Compliance
Sources
This page was written from 5 sources, 3 on domains other than salesforce.com.
- 1.salesforce.com — claudeforcevendor
- 2.venturebeat.com — salesforce just put its entire crm inside claude and says yo
- 3.salesforceben.com — salesforce and anthropic announce claudeforce in q2 27 earni
- 4.salesforcetime.com — claudeforce salesforce meets claude
- 5.salesforce.com — salesforce and anthropic announce claudeforcevendor
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
Aziron
Self-hosted enterprise agent execution platform that turns AI answers into governed actions in the system of record
ChipAgents
Agentic AI for RTL design, verification and waveform debug
AccuKnox AgentZ
Build, run and govern AI agents with sandboxing and tool-level permissions underneath
HappyRobot
AI workers that run enterprise voice, email and document operations end to end