AI Services · Explainer
Claudeforce: Claude in Salesforce, Agentforce and Slack
Salesforce and Anthropic announced Claudeforce in August 2026. Here is what it includes, what is available now, and how to decide where Claude should run in your Salesforce org.
Short answer
Claudeforce is the Salesforce and Anthropic partnership announced on August 26, 2026. It covers four things: the Salesforce in Claude plugin with prebuilt sales skills, Claude as a reasoning model inside Agentforce through Amazon Bedrock within the Salesforce trust boundary, Claude as the default model in Slack, and AIforce, Salesforce's layer that exposes CRM data and actions to agents through MCP, APIs and the CLI. Parts were in pilot or beta at launch, so confirm availability for your org.
Until this year the question was Agentforce or Claude. Claudeforce makes it a question of where each piece of work should run. A seller who lives in Claude can pull live pipeline and take governed actions through the Salesforce in Claude plugin. A service agent built in Agentforce can use Claude for its reasoning while Salesforce keeps the topics, actions, Einstein Trust Layer controls and audit. Slack users get Claude as the model behind Slackbot.
For regulated firms the important detail is the trust boundary. Claude in Agentforce runs through Amazon Bedrock inside Salesforce's boundary, which keeps model calls under the controls your Salesforce contract already covers. Claude connected directly through MCP runs under Anthropic's commercial terms and your Claude admin settings. Both can be right. They need different reviews.
What Claudeforce includes
| Component | What it does | Status |
|---|---|---|
| Salesforce in Claude | Plugin with 37 prebuilt sales skills (meeting prep, deal health, pipeline review) that read Salesforce data and route actions back through Salesforce business rules | Select pilot customers at launch; Salesforce now lists a beta sign-up |
| Claude in Agentforce | Claude as the reasoning model for the Atlas Reasoning Engine and in Agent Builder, through Amazon Bedrock inside the Salesforce trust boundary | Announced; confirm your org and edition with Salesforce |
| Slack | Claude as the default model for Slackbot and related Slack AI features | Announced |
| AIforce | Salesforce's enterprise layer that delivers data and workflows to agents through MCP servers, APIs and CLI tools | Announced |
Product names, availability and licensing come from Salesforce and Anthropic. We check them for your org before we scope work.
What you get
- A placement map: which use cases run in Claude, which in Agentforce, which in Slack, and why
- Licensing and consumption review with your Salesforce and Anthropic contracts
- Agentforce agent design with Claude as the model where your org supports it
- Salesforce in Claude plugin rollout for a pilot sales team once it is available to you
- Permission sets, integration users and External Client Apps scoped per path
- Evaluation sets and monitoring that work across both surfaces
How it comes together
- Inventory the work. List the jobs you want AI to do and who does them today.
- Place each job. Inside-Salesforce agents go to Agentforce; cross-tool assistant work goes to Claude; team conversation goes to Slack.
- Confirm licensing. Agentforce consumption, Claude seats and API usage are billed differently; we model the cost before build.
- Build and test. One agent or one skill set at a time, scored against real records.
- Govern both sides. One register of AI tools, one review cadence, one owner per use case.
Is this the right call?
Good fit when
- You run Salesforce and want Claude's reasoning without leaving the Salesforce trust boundary
- Sellers already use Claude and want live pipeline and governed updates in the same place
- You are planning Agentforce and want a second opinion on model choice
Think twice when
- You need production today and your org isn't in the pilot or beta; use hosted MCP now and move later
- Your main CRM is HubSpot; see Claude + CRM integration and Breeze
Mistakes we help you avoid
Treating every path as the same review
Bedrock inside the Salesforce boundary and direct Anthropic access fall under different contracts and settings. Review each one.
Buying ahead of availability
Pilot and beta features change. Scope the first phase on what your org can turn on today.
Questions we get
What is Claudeforce?
Claudeforce is the partnership Salesforce and Anthropic announced on August 26, 2026. It brings Salesforce data and actions into Claude through the Salesforce in Claude plugin, makes Claude available as a reasoning model in Agentforce through Amazon Bedrock inside the Salesforce trust boundary, and makes Claude the default model in Slack.
Should we use Agentforce, Claude, or both?
Use Agentforce for agents that live inside Salesforce channels and depend on Salesforce topics, actions and the Einstein Trust Layer. Use Claude directly for cross-tool assistant work, long documents and custom applications. Many firms use both, and Claudeforce lets Agentforce agents run on Claude where supported.
Is the Salesforce in Claude plugin available now?
At announcement it was available to select pilot customers, and Salesforce now lists a beta sign-up, with more skills planned for late 2026. We confirm what your org can enable before scoping. Salesforce hosted MCP servers are a generally available alternative today.
Does Claude in Agentforce cost extra?
Claude in Agentforce is contracted through Salesforce, so consumption follows your Agentforce agreement rather than Anthropic pricing. Ask your Salesforce account executive how it is metered for your org; we model the expected usage before build.