# Claudeforce

> Source: https://aiwiki.ai/wiki/claudeforce
> Updated: 2026-08-27
> Fact-checked: 2026-08-27
> Categories: AI Agents, AI Tools & Products, Anthropic, Enterprise AI
> License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) - attribute to "AI Wiki (aiwiki.ai)"
> Cite as: AI Wiki. "Claudeforce." aiwiki.ai, 27 Aug 2026. https://aiwiki.ai/wiki/claudeforce
> From AI Wiki (https://aiwiki.ai), the free encyclopedia of artificial intelligence. Reuse freely with attribution.

**Claudeforce** is an expanded partnership and enterprise integration portfolio announced by [Salesforce](https://aiwiki.ai/wiki/salesforce) and [Anthropic](https://aiwiki.ai/wiki/anthropic) on August 26, 2026. It connects Salesforce data, workflows, permissions, and business logic with [Claude](https://aiwiki.ai/wiki/claude). Claudeforce is not an AI model or a replacement for either company's main platform. The name covers the partnership and the products announced under it.[1][2]

The first product is **Salesforce in Claude**, a plugin with 37 prebuilt sales skills. At announcement, the plugin was available only to select pilot customers. The companies expected an open beta in September 2026 and said that more skills would begin arriving later in 2026.[1][2] These dates were plans rather than completed releases.

Claudeforce also covers the use of Claude within Salesforce products, expanded Claude integration across Slack, and reciprocal adoption of each company's software. This makes it broader than Salesforce in Claude, which is one product for working with Salesforce from the Claude interface.[1]

## Key facts

| Field | Detail |
| --- | --- |
| Announced | August 26, 2026[1][4][5] |
| Companies | Salesforce and Anthropic[1] |
| Subject | Strategic partnership and enterprise integration portfolio[1][2] |
| First product | Salesforce in Claude[1][2] |
| Initial users | Sales and revenue teams[1][2] |
| Initial capability set | 37 prebuilt sales skills[1][2] |
| Launch access | Select pilot customers[1][4][5] |
| Planned next stage | Open beta in September 2026[1][2] |
| Connection layer described by Salesforce | AIforce and Headless 360, using MCP servers, APIs, and command-line tools[1][2][3] |
| Related Salesforce platform | [Agentforce](https://aiwiki.ai/wiki/agentforce), where Claude is available as a reasoning model[1] |
| Public evidence at announcement | Vendor documentation and independent reporting; no product-specific independent security or performance evaluation located |

## Background and announcement

Salesforce and Anthropic had integrations before the Claudeforce name was introduced. The companies had made Claude available in Salesforce products, worked on Model Context Protocol connections, and integrated Claude with Slack. Salesforce's announcement presented Claudeforce as the framework that brought those efforts into one expanded partnership.[1][2]

Patrick Stokes, Salesforce's president of Applications and Marketing, told CIO that the companies began productizing the integration after users connected Salesforce to agents through Headless 360. According to Stokes, developer-built connections became difficult to operate consistently for hundreds or thousands of users, particularly around managed authentication and user permissions. Anthropic's sales organization had encountered similar scaling problems while using Salesforce from Claude.[4] This account explains the product's origin but comes from a Salesforce executive rather than an independent technical study.

CNBC reported that the announcement was the first time Salesforce had attached its "force" suffix to another company's product name. The report also confirmed that the initial release was a limited pilot, not a generally available product.[5]

## Relationship between the products

Claudeforce, Salesforce in Claude, Agentforce, Headless 360, and AIforce refer to different parts of the integration.

| Name | Role in the announcement |
| --- | --- |
| Claudeforce | The Salesforce-Anthropic partnership and the portfolio of integrations announced under it[1][2] |
| Salesforce in Claude | The first Claudeforce product, a plugin that lets permitted users work with Salesforce data and actions from Claude[1][2] |
| Agentforce | Salesforce's existing [AI agent](https://aiwiki.ai/wiki/ai_agents) platform; Claude is available within it as a reasoning model[1] |
| Headless 360 | Salesforce architecture that exposes platform capabilities through APIs, MCP tools, and command-line tools[3] |
| AIforce | Salesforce's term for an enterprise harness that supplies agents with business data and workflows through MCP servers, APIs, and command-line tools[1] |

The direction of integration matters. Salesforce in Claude brings Salesforce capabilities into Claude. Claude inside Agentforce brings Anthropic's model into Salesforce's agent platform. Both are part of the wider partnership, but one is not a renamed version of the other.[1]

## Salesforce in Claude

Salesforce in Claude was introduced for sellers and revenue teams. Its 37 skills include meeting preparation, deal health review, and pipeline review. Salesforce said the skills were built jointly with Anthropic for Claude's reasoning, tool use, and generated interfaces rather than as a set of generic prompts.[1]

The plugin is designed to read live Salesforce context that a user is allowed to access, answer questions about accounts and pipeline, and submit permitted changes back to Salesforce. The announcement says an onboarding flow can use context from Salesforce, [Slack](https://aiwiki.ai/wiki/slack), and other connectors enabled in Claude to generate a tailored sales dashboard. When the plugin takes an action, Salesforce says it routes that action through Salesforce so existing business rules apply.[1][2]

Salesforce describes setup as a central administrative connection. Authentication and permissions are managed for the organization, while individual actions use the permissions of the Salesforce user. The product page also says customers control how much authority Claude receives for write operations and when it may update Salesforce.[2] These are vendor-described controls. The launch materials did not include an independent security assessment or a complete public matrix of the permissions required by each skill.

CIO reported that use of the plugin consumes model tokens. Stokes said early usage was well below the token consumption of software-development workloads, but the companies did not publish a measured comparison or final pricing model with the announcement.[4] The launch pages also did not specify the exact Salesforce editions, Claude plans, regions, or contractual terms that would apply after the pilot.

## Architecture and permission model

Salesforce says Salesforce in Claude uses Headless 360 to expose Salesforce data, workflows, business logic, and permissions through a secure [Model Context Protocol](https://aiwiki.ai/wiki/model_context_protocol) connection. Its broader Headless 360 architecture exposes platform capabilities as APIs, MCP tools, and command-line tools, while retaining Salesforce identity, access controls, and governance. The Claudeforce page says MuleSoft manages access at scale.[2][3]

The formal announcement also names AIforce as the Salesforce harness beneath Salesforce in Claude. It describes AIforce as a layer that makes business data and workflows available to agents through MCP servers, APIs, and command-line tools. The launch materials mention AIforce and Headless 360 as related parts of the stack, but do not define them as interchangeable names.[1]

MCP standardizes communication between a model-based application and external data or tools. Its HTTP authorization specification defines OAuth-based client, server, and resource-owner roles, token audience validation, and scope selection. Authorization remains optional at the protocol level, and the specification recommends that clients request only the scopes needed for an operation.[6] The use of MCP therefore does not, by itself, verify Salesforce's product-specific claims. The effective boundary depends on the Salesforce and Claude implementations, their identity configuration, scopes, business rules, logging, and administrative settings.

Academic work on MCP has cataloged risks that can arise in tool-connected agent systems, including prompt injection, tool poisoning, and unauthorized operations. A separate 2026 interview study of 20 practitioners at eight companies found that enterprise users valued cross-system coordination but still encountered problems with distributed state and fault diagnosis.[7][8] Neither study evaluated Claudeforce, Salesforce in Claude, Headless 360, or Salesforce's MCP servers. They provide general context for why concrete implementation and operational evidence matter, not evidence of a flaw in this product.

## Other parts of the partnership

The announcement describes three areas beyond the initial sales plugin. First, Claude is available inside Agentforce as a reasoning model for the Atlas Reasoning Engine. Salesforce also identified Claude as the default model for Agentforce Vibes and Agentforce Coworker and as an option in Agent Builder.[1]

Second, Salesforce and Anthropic expanded their work in Slack. The companies said Claude would be the default model for Slack and would support Slackbot, Claude Tag, and Slack Code. These integrations place Claude in a Salesforce-owned work surface while Salesforce in Claude moves in the opposite direction, placing Salesforce capabilities in Anthropic's interface.[1]

Third, each company identified the other as a strategic customer. Salesforce said it was Anthropic's preferred customer relationship management provider and that Slack was Anthropic's preferred internal team platform. Salesforce named Claude as the default model or preferred assistant for several products and employee groups, and said it planned to make Claude Code and Claude Enterprise available to its developers and knowledge workers.[1] The announcement did not disclose the financial terms of these arrangements.

## Availability and evidence status

Salesforce in Claude was in a select-customer pilot on August 26, 2026. Salesforce's product page listed the open beta for September and labeled capabilities for service, marketing, commerce, revenue management, field service, Tableau, MuleSoft, Informatica, Data 360, Headless 360, and industry products as coming soon.[2] Those entries were a roadmap, not evidence that every listed integration was available.

The detailed architecture, permission behavior, and deployment claims at launch came from Salesforce and Anthropic. CIO and CNBC independently confirmed the announcement, pilot status, product identity, and release plans, but did not conduct a technical audit.[4][5] The public launch materials did not include a product-specific benchmark, penetration-test report, complete threat model, availability service level, or generally available pricing. Later beta documentation, customer studies, and independent security testing would provide evidence that was not available at announcement.

## References

1. Salesforce. "Salesforce and Anthropic Announce Claudeforce: The #1 AI Meets the #1 AI CRM." Aug. 27, 2026 page date; San Francisco dateline Aug. 26, 2026. https://www.salesforce.com/au/news/press-releases/2026/08/27/salesforce-and-anthropic-announce-claudeforce/
2. Salesforce. "Claudeforce: The #1 AI Meets the #1 CRM." Product page and FAQ, accessed Aug. 27, 2026. https://www.salesforce.com/claudeforce/
3. Salesforce. "Salesforce Headless 360: Headless Agentic AI CRM." Product and architecture page, accessed Aug. 27, 2026. https://www.salesforce.com/headless/
4. Thor Olavsrud. "Salesforce, Anthropic partner to deliver Claudeforce." CIO, Aug. 26, 2026. https://www.cio.com/article/4214458/salesforce-anthropic-partner-to-deliver-claudeforce.html
5. Ashley Capoot. "Salesforce, Anthropic expand partnership as Benioff responds to 'SaaSpocalypse' concerns." CNBC, Aug. 26, 2026. https://www.cnbc.com/2026/08/26/salesforce-anthropic-partnership-claudeforce.html
6. Model Context Protocol. "Authorization." Protocol revision 2026-07-28, accessed Aug. 27, 2026. https://modelcontextprotocol.io/specification/2026-07-28/basic/authorization
7. Shiva Gaire et al. "Systematization of Knowledge: Security and Safety in the Model Context Protocol Ecosystem." arXiv:2512.08290, revision 2, Dec. 13, 2025. https://arxiv.org/abs/2512.08290
8. "Understanding How Enterprises Adopt the Model Context Protocol for LLM-Driven Software Engineering." arXiv:2606.09182, 2026. https://arxiv.org/abs/2606.09182