Salesforce Inc. has introduced a collection of new tools designed to support customers’ artificial intelligence initiatives.

The first set of features, which debuted today, will help organizations manage the data they use to power their AI models. Salesforce says that the new capabilities make it easier to secure AI models’ data and ensure their prompt responses are reliable. The enhancements are rolling out alongside Agentforce Vibes, a vibe coding tool that was announced on Wednesday.

Simplified data governance 

Most of Salesforce’s new data management features are rolling out for its Data Cloud service. It’s a customer data platform, or CDP, that enterprises can use to organize sales and marketing information from different sources in a centralized repository. Storing information in one place makes it easier to fix errors and run analyses.

The first new feature that debuted today is called Context Indexing. According to Salesforce, it helps AI agents interpret unstructured data such as contracts and schematics. Context Indexing enables users to upload an unstructured file and have an AI agent generate a detailed natural language explanation of its contents.

The feature is joined by new AI-powered cybersecurity capabilities designed to help companies protect the data they keep in Salesforce. The additions include integrations with CrowdStrike Holdings. Inc and Okta Inc. products. The two tech firms’ applications are used for tasks such as detecting cyberattacks.

Salesforce detailed the feature additions in conjunction with the launch of a previously announced capability, Data Cloud Clean Rooms, into general availability. The tool is geared toward organizations that require the ability to share sensitive internal data with third parties. According to Salesforce, Data Cloud Clean Rooms makes it possible to share information without creating additional copies that can be misplaced.

Rounding out the list of data governance upgrades is a feature called Customer 360 Semantic Data Model. According to Salesforce, it can help companies ensure that the metrics they use to measure their go-to-market activities are consistent. 

Vibe coding

Agentforce Vibes, the new vibe coding tool that Salesforce debuted on Wednesday, uses large language models to automate manual programming tasks. It works with VS Code, a popular open-source code editor. Agentforce Vibes also supports third-party distributions of the editor.

The tool can generate code using GPT-5 and Salesforce’s internally developed xGen series of open-source LLMs. Additionally, customers can connect it to internal LLMs they host on their own infrastructure.

Salesforce says that Agentforce Vibes can generate code, test it and automatically fix the bugs it finds. The tool can use Salesforce’s Apex programming language to build applications that run on the cloud giant’s platform. Agentforce Vibes also supports other syntaxes including HTML and CSS, which are geared toward interface development.

Salesforce says that competing vibe coding tools are suitable for prototyping tasks, but often struggle to generate production-grade code. According to the company, the reason is that rivals lack production-grade cybersecurity and regulatory compliance features. It claims that Agentforce Vibes fares better in that department.

The tool is compatible with Salesforce Sandboxes, which are isolated cloud environments that developers can use to test AI-generated code. The feature makes it possible to search for cybersecurity issues before a workload rolls out to production. Agentforce Vibes also finds other problems such as performance bottlenecks. 

Software teams that adopt the tool can use Trust Layer, a component built into Salesforce’s platform, to prevent their applications’ LLMs from generating harmful output. The module filters prompt responses that contain data such as credit card numbers. 

Agentforce Vibes can tailor the code it outputs to a company’s Salesforce environment. It does so by taking into account the environment’s schema, a file that describes how business records are formatted. That capability and several of Agentforce Vibes’ other features are powered by an artificial intelligence agent dubbed Vibe Codey. 

“Vibe Codey is context-aware, meaning it has a deep understanding of your Salesforce project structure and ‌organization’s metadata,” Dan Fernandez, Salesforce’s vice president of products for developer services, wrote in a blog post. “Vibe Codey can discover, analyze and reuse existing code, and adhere to coding standards for greater consistency and collaboration.”

On launch, Agentforce Vibes supports a “limited number” of prompts per user. Salesforce plans to roll out a feature that will enable customers to raise the tool’s usage caps for a fee. Additionally, the cloud giant will introduce support for more additional coding models.

Agentforce Vibes is the latest fruit of a years-long effort by Salesforce to integrate more AI capabilities into its platform. Previously, the company introduced Agentforce 3, a new version of its flagship AI feature set. The release brought improved MCP support, capabilities for monitoring AI agents’ performance and other enhancements. 

Photo: Salesforce