AI

Salesforce Opens 'Claudeforce' Beta — What It Actually Ships

Live CRM data inside Claude, 37 prebuilt sales skills, and a 12% stock pop. Here's what teams actually get.

Mara Oduya

Senior AI Correspondent

Published 3 min read
Contemporary workspace featuring computers, coding screens, and office essentials in a tech environment.
In this article 7 sections

Quick answer: Salesforce and Anthropic announced Claudeforce, putting live CRM data and governed actions directly inside Claude. It ships with 37 prebuilt sales skills, an open beta for eligible customers was planned for September, and Salesforce stock jumped 12% the day the expanded partnership was announced. Here's what teams actually get, not the press-release version.

This is the first time Salesforce has put its "force" suffix on another company's product. That alone tells you how this deal got internally prioritized. The product itself is more interesting than the branding decision, so let's get into it. At Protocol Wire, we cover AI launches by what they ship.

What's Actually in the Beta

Contemporary workspace featuring computers, coding screens, and office essentials in a tech environment.

Claudeforce's first release, called Salesforce in Claude, ships with a plugin containing 37 prebuilt sales skills, according to Salesforce's own announcement. Sellers and their agents can pull live revenue context, update a pipeline, and take governed actions directly from inside Claude instead of switching to the CRM. Select pilot customers have had it since August; the open beta for a broader customer base was planned to start in September, with more skills promised through the rest of the year. If you've used any "AI meets CRM" tool before, the pitch sounds familiar. What's different here is the data access — this isn't a chatbot summarizing your CRM from the outside, it's Claude with the same live pipeline data your reps already see.

How the Plugin Is Set Up and Governed

The 37 skills cover tasks such as meeting prep, deal health review, and pipeline review. Per Salesforce's announcement, onboarding reads a seller's connected context, including Salesforce, Slack, and other Claude connectors, and builds a tailored dashboard with interactive views. Actions route through Salesforce, so existing business rules still apply.

Administration is meant to be a one-time job. An admin connects Salesforce in Claude once, with authentication and permissions managed centrally, and every seller on the team gets access without per-user setup or a new permissions model. The plugin runs on AIforce, which Salesforce describes as its enterprise harness for connecting business data and workflows to agents through MCP servers, APIs, and CLI tools.

The write access is the part buyers will scrutinize. The Next Web's analysis noted that read-only agents are easy to demo, while write access requires permissions, audit trails, and revocation. Channel Insider also reported that the two companies are developing what they call "Enterprise Frontier Safeguards" to protect Salesforce data and keep model behavior within defined boundaries.

Why Salesforce Put Its Name on Someone Else's Product

Group of professionals discussing charts during a business meeting.

Salesforce shares rose 12% in extended trading the day this was announced, alongside a quarter that beat Wall Street's expectations, per CNBC's coverage of the announcement. Read that stock move for what it is: the market rewarding Salesforce for admitting a competitor's model is good enough to build the company's own product around, instead of insisting on an in-house alternative nobody was asking for. That's a real strategic call, not a marketing stunt — Benioff's been fielding "does AI make Salesforce's core product obsolete" questions for two years, and this is the company's answer.

What This Costs Sales Teams to Adopt

A tablet with a financial website open, alongside a white keyboard and mouse. Ideal for tech-theme stock images.

The beta itself is free to eligible existing customers right now, but "free beta" and "free forever" are two different sentences, and nobody's published what the eventual pricing tier looks like once this graduates out of beta. Budget for it the way you'd budget for any new enterprise AI seat license, not the way you'd budget for a feature bundled into your existing contract. If you want the fuller picture on why AI infrastructure costs tend to arrive as a surprise line item rather than the number on the pricing page, this site's breakdown of what running AI workloads actually costs is worth a read before your renewal conversation, not after.

Where Claude Sits Inside Salesforce

Claudeforce is broader than the plugin. According to Salesforce, Claude is now a reasoning model for the Atlas Reasoning Engine and the default model for Agentforce Vibes and Agentforce Coworker, and it is available in Agent Builder. In Slack, Claude is the default model for Slackbot, Claude Tag, and Slack Code.

For regulated industries, Salesforce says Claude is reachable through Amazon Bedrock within the Salesforce Trust Boundary. The release also says Slackbot has produced 8.1 million hours of annualized productivity gains for Salesforce employees, a figure up more than 2x quarter over quarter. That is a company-reported internal number, not a customer result.

The financial ties are deep. The Next Web reported that Salesforce expects to spend about $300 million on Anthropic tokens this year and holds a stake in Anthropic valued at roughly $5 billion. The same report flagged the dependency risk of making one supplier's model the default across a product line.

On cost structure, Channel Insider's coverage said customers pay Salesforce for the underlying data and services while Anthropic supplies the model under token-based inference. It also said Salesforce must show agent-driven usage creates enough value to offset pressure on traditional per-seat pricing. Neither outlet reported Claudeforce pricing.

What to Watch Next

Close-up of two businessmen shaking hands, symbolizing agreement and partnership.

Two things worth tracking. First, whether the governed-action piece actually holds up under a real sales team's messy day-to-day use, or whether it turns into another pilot that stalls the way most enterprise AI agent pilots do once it hits a real approval process. Second, whether Salesforce's own buyers are actually ready for it — plenty of enterprise software buyers are still behind on agentic AI readiness, and a good product doesn't fix an unready buyer.

Conclusion

A diverse group of professionals working together in a modern conference room with laptops.

Claudeforce is a real product with real data access, not a rebrand. Whether it's worth adopting on day one of a beta depends entirely on whether your team can absorb a new AI seat into the workflow without it becoming shelfware by Q2. Watch the pricing announcement and the first round of customer case studies before committing budget — the press release already told you what it does. It hasn't told you what it costs. Protocol Wire will keep covering the cost side of enterprise AI.

What is Claudeforce?
An expanded partnership between Salesforce and Anthropic that puts live CRM data, workflows, and governed actions directly inside Claude, announced in August 2026 as Salesforce in Claude, with an open beta planned for September 2026.
What does the Claudeforce beta actually include?
A plugin with 37 prebuilt sales skills that let sellers and agents reason over live revenue data, automate pipeline updates, and take governed actions from inside Claude.
How much does Claudeforce cost?
Salesforce hasn't published post-beta pricing. It's free to eligible existing customers during the open beta, but that's not the same as free once it moves to general availability.
Why did Salesforce put its "force" branding on an Anthropic product?
It's the first time Salesforce has used that naming convention for another company's product, signaling the partnership is a core strategic bet rather than a feature integration — reinforced by a 12% rise in Salesforce shares in extended trading the day it was announced.

Written by

Mara Oduya

Senior AI Correspondent

9+ years experience

Before joining Protocol Wire, Mara spent five years building ML infrastructure at an ad-tech company, which is the polite way of saying she has personally paged herself at 3 a.m. over a feature store outage. She writes the AI beat with a builder's skepticism: she wants to see the eval, not the demo. Her reporting has traced how agent frameworks actually get adopted (or quietly ripped out) inside real engineering orgs, and she keeps a running, only-half-joking tally of AI product launches that walk back their own benchmark claims within a quarter.

Credentials

  • B.S. Computer Science, Georgia Institute of Technology
  • Former Machine Learning Infrastructure Engineer, adtech sector
  • 2021 Data & Society journalism fellow