Agentic artificial intelligence runs on data, so Salesforce Inc. today announced it is launching Headless Data 360 for Model Context Protocol to provide that data directly to agents, allowing them access to relevant, governed customer context.

Teams can do more than query data through Data 360. Now they can build, transform, map, segment and activate fields as well – all without leaving existing tools and interfaces.

That’s because Data 360 exposes more than 200 existing application programming interfaces as programmable endpoints that AI agents can use without being locked behind a user interface. This means they can be used by a machine directly, instead of needing specialized tools or a human to sit down with a book or a search engine, a soda and a lot of extra time.

Historically, working with data has been difficult because data workers had to know what fields, schema, segments and tables they were calling before activating them in a database. This meant they had to constantly move between tools and teams to pull in the data they needed, which could take days or weeks to finalize a project. Now, with Data 360 and access to all the APIs in one place, this can happen all at once.

In April, Salesforce released Headless 360, a new method for connecting any system on Salesforce APIs via MCP, which allows AI agents to talk to each other and data sources.

The company described this as the “front door” to the entire ecosystem, allowing developers to build a conversational interface for any part of Salesforce. The new Data 360 goes deeper, providing deep intelligence, insights, transformation and ingestion for data through AI agents.

That opens up new paths for a business team to build data paths to insights from natural language by describing what’s needed — for example, “lifetime value of all customers for electronics purchases, excluding software.” Then it can have the agent write the query, which creates the calculation and runs it. It will construct the semantic models, relationships, transforms and formulas, all hands-off.

That means a user can use any AI agent from Claude, Cursor, ChatGPT or Salesforce’s own Agentforce, and it will dynamically discover, understand and invoke the necessary Salesforce data and capabilities, without being told exact field names or function calls. Since all the APIs are available, the developer doesn’t need to entertain the agent with precision, just the concept in their head and the context of their project, allowing them to ideate and iterate like talking to a teammate.

Conversational interfaces can build data transforms, get automated intelligent field-mapping suggestions across different systems, query and inspect data graphs, down to individual records. It can also interrogate new data streams coming from a brand-new organization’s customer relationship management system to help prepare its data as it’s being brought into Data 360. It takes what might have taken months of backbreaking work and turns it into hours or minutes instead.

“Data 360 Headless lets our agents reach unified patient context from any surface — through MCP, without waiting for a purpose-built interface for each one,” said Dr. Paramjit “Romi” Chopra, founder and chief executive of Midwest Institute for Minimally Invasive Therapies P.C.

He explained that what makes agent answers more trustworthy from Data 360 is that the metadata is “physician-curated ontology,” meaning the responses are structured by humans; even if a generative AI machine is responding, it’s being monitored by experts.

Salesforce added that Data 360 MCP Server is Salesforce-hosted and managed and remains consistent with every other service and MCP server. It comes with several prepackaged Skills for common tasks such as modeling, mapping, transforms and activation.

Initial agent Skills are designed to handle tedious labor out of the box, eliminating the need for users to build or program these capabilities themselves. Custom Skills can also be used to encapsulate common workflows and prompts that users request agents complete consistently, saving users both time and effort. Skills will be available later in August.

Image: Nepool/Shutterstock