What Is Salesforce Headless 360? The AI Agent Platform Explained

Salesforce just made its entire platform usable without a browser. At TDX 2026, the company unveiled Salesforce Headless 360 — an architectural overhaul that exposes every Salesforce capability as an API, an MCP tool, or a CLI command. AI agents, scripts, and third-party software can now operate Salesforce without a human ever logging in.

This is being called the most significant platform shift in Salesforce's 27-year history. Here's what it actually means.

What Is Salesforce Headless 360?

Salesforce Headless 360 is an API-first architecture that makes the entire Salesforce platform — data, workflows, business logic, and compliance controls — accessible to AI agents and developers without a graphical user interface.

It was officially announced on April 15, 2026, at TrailblazerDX 2026 (TDX 2026) in San Francisco. Salesforce states the underlying rebuild began approximately two and a half years prior.

At launch, Salesforce grouped the announcement into three pieces: new MCP tools and coding skills, a new Agentforce Experience Layer that renders native interactions across different surfaces, and new control tools for evaluating and governing agent behavior in production. The launch shipped over 100 new developer tools, including more than 60 new MCP tools and 30 preconfigured coding skills empowering coding agents with complete, live access to your entire platform. This includes all of your data, workflows, and business logic, directly in the coding agents you might already use, such as Claude Code, Cursor, Codex, and Windsurf.

Salesforce says these capabilities now span the company's system of context (Data 360), system of work (Customer 360), system of agency (Agentforce), and system of engagement (Slack). (salesforce.com)

The simplest way to think about it: Salesforce used to be a place humans went to do work. Now it's infrastructure that agents call to get work done.

Why "Headless"? What the Term Actually Means

In software, "headless" means removing the visual interface — the "head" — and exposing functionality through programmatic endpoints instead. A headless CMS stores and serves content without dictating how it's displayed. A headless browser renders web pages without showing them on screen.

Salesforce Headless 360 applies the same concept to an entire enterprise platform. Headless removes that assumption by exposing the capabilities that sit behind those applications directly — which means data can be retrieved, actions invoked and processes triggered without going through the interface.

The key shift: instead of designing for a person clicking through screens, Salesforce is re-optimizing its platform for agents, not people.

The 3 Ways to Access Salesforce Now: APIs, MCP, and CLI

Headless 360 gives agents and developers three distinct access channels. They overlap, but they serve different purposes.

Access path	Best for	What changes with Headless 360
APIs	Custom apps, deterministic integrations, back-end services	Salesforce capabilities are easier to expose and compose for software-first workflows
MCP tools	AI agents, copilots, ChatGPT or Claude-style clients	Agents can discover and invoke curated tools instead of relying on brittle prompt hacks
CLI / DevOps	Scripts, CI/CD, admin and dev release work	Dev and release tasks can be driven from local tooling, pipelines, and natural-language DevOps

APIs

APIs remain the right tool when you want precise, deterministic application behavior. Headless 360 exposes Salesforce's underlying data, workflows, and governance controls as APIs, MCP tools, and CLI commands. Salesforce's API Catalog lets admins expose selected REST endpoints as named tools, but each API Catalog tool maps to a single REST endpoint, and not every endpoint is cataloged at general availability. If the workflow needs multi-step logic, Salesforce points you to Flows or Invocable Actions rather than pretending one endpoint will do everything. (developer.salesforce.com)

MCP Tools

MCP (Model Context Protocol) is the piece most people actually mean when they talk about Headless 360. MCP is an open-source standard, originally created by Anthropic, that provides a universal way for AI agents to connect to external tools and data. MCP is a protocol drafted by Anthropic and has emerged as the de facto standard for describing how an agent should access tools, data sources and systems. The usual shorthand is "USB-C for AI."

Salesforce Hosted MCP Servers let compatible clients — Claude, ChatGPT, Cursor, or a custom agent — connect through OAuth and act on behalf of an authorized user. The DX MCP Server includes over 60 MCP tools, organized into toolsets so developers can scope access and avoid overwhelming an LLM's context window. Prebuilt servers cover SObject access, Data 360 SQL, Flows, Invocable Actions, Prompt Builder, Tableau Next, and API Catalog endpoints. (developer.salesforce.com)

The security model matters here. Hosted MCP Servers enforce per-user authentication and honor object permissions, field-level security, and sharing rules on every tool call. Salesforce ships read-only, mutation-only, delete-only, and full-access SObject servers — a practical way to scope what an agent can do before you hand it write access. (developer.salesforce.com)

Salesforce's own docs point cautious teams to the read-only platform/sobject-reads server first. That's smart advice. Read-only access is enough for Q&A, summaries, reports, prep work, and human-in-the-loop review — and it removes the risk of silent record changes while you learn how your agent actually behaves. (developer.salesforce.com)

CLI and DevOps

The Salesforce CLI lets agents and scripts run deployment, testing, and org-management tasks from a terminal. Salesforce's DX MCP Server is a local, open-source MCP server built by the DX team — not just a wrapper around shell commands. It integrates with DX Foundation libraries and exposes toolsets for orgs, metadata, users, data, and logic. (developer.salesforce.com)

Combined with the DevOps Center MCP, developers can describe a deployment in natural language and let an agent execute it. The build loop that used to require context-switching across four different tools now happens inside one connected experience — cutting cycle times by as much as 40%.

All three channels inherit existing Salesforce permissions, sharing rules, and compliance controls. No separate trust layer to rebuild.

Why This Matters: The End of the Browser Login

Enterprise SaaS has been built on a single assumption for decades: a human logs in, navigates a UI, and clicks things. Historically, Salesforce functionality has been consumed via applications. Users log in, navigate screens and follow predefined flows.

Headless 360 breaks that assumption. It treats non-human operators as first-class users.

Salesforce Co-Founder Parker Harris framed the logic bluntly before the launch: "Why should you ever log into Salesforce again?"

This isn't a theoretical exercise. Travel platform Engine, featured in the TDX keynote, built its customer service agent in 12 days using Agentforce and now handles 50% of customer cases autonomously. The agent operates through Slack and API calls — no CRM tab open, no human clicking through case records.

The shift isn't just automation. It's how much of the platform Salesforce is now explicitly packaging around agentic AI as a first-class runtime model, with the Experience Layer separating what an agent does from where the interaction shows up. The interface is no longer the product. The data, workflows, and business logic underneath are the product. The surface where interactions happen — Slack, Teams, WhatsApp, ChatGPT, a custom React app — becomes interchangeable. The surface changes. The platform doesn't.

How Headless 360 Fits With Agentforce

This is where most of the confusion sits, so let's be precise.

Agentforce is Salesforce's own AI agent system — the product you use to build, deploy, and manage AI agents inside the Salesforce ecosystem. It's the "system of agency."

Headless 360 is the underlying infrastructure that opens up Salesforce to any agent — Salesforce's or third-party. It's the plumbing.

Salesforce has organized its platform around four layers: a system of context (Data 360), a system of work (Customer 360 apps), a system of agency (Agentforce), and a system of engagement (Slack and other surfaces). Headless 360 opens every layer via programmable endpoints.

The better mental model is not competition but stack order. Agentforce runs on top of Headless 360-style access patterns, and external tools can too. Salesforce Hosted MCP Servers already support third-party clients like ChatGPT, Claude, and Cursor — which is why Headless 360 matters even if you never plan to use Salesforce's own agent UX. (developer.salesforce.com)

The Agentforce Experience Layer — a new UI service shipped with Headless 360 — separates what an agent does from how it appears, rendering rich interactive components natively across Slack, mobile apps, Microsoft Teams, ChatGPT, Claude, Gemini, and any client supporting MCP apps.

Build your agent logic once, deploy it across every surface.

What It Means for Businesses and Data Readiness

For business teams, the practical impact is faster work happening closer to the system of record. A service agent can pull account context, inspect case history, trigger a flow, and return the result in Slack without anyone opening a Salesforce console. A developer can query org data, deploy metadata, or run tests from an MCP-enabled IDE.

Here's where the trade-offs get real:

• Automation goes deeper. Agents no longer just surface information — they can execute workflows, trigger approvals, and update records end-to-end without a human in the loop.
• Agents inherit your business logic. When an agent operates inside Salesforce, it inherits approval chains, business rules, and edge-case logic built years ago, and does not have to rediscover or approximate them.
• Permissions become product behavior. If an agent acts as the user, weak security design turns into weak agent behavior. Hosted MCP calls inherit the same field, object, and sharing rules as the connected user. (developer.salesforce.com)
• Tool curation matters. Salesforce explicitly warns that AI clients struggle when you dump too many tools into a single surface; focused, persona-specific servers work better. (developer.salesforce.com)
• Multi-model flexibility. The platform now integrates with OpenAI, Anthropic, Google Gemini, Meta's LLaMA, and Mistral AI models. You're not locked into a single LLM vendor.
• Not every feature is equally mature. Features are arriving in phases, and analysts expect teams to supplement early governance features with their own evaluation frameworks for a while. (cio.com)

But there's a catch that doesn't show up in keynote demos: AI agents are only as good as the data they operate on. An agent calling an MCP tool to check a customer's renewal status will return garbage if the underlying Salesforce data is duplicated, unmapped, or stale. An integration that feeds pipeline data from HubSpot into Salesforce has to be airtight — because now an autonomous agent, not a human who might catch the error, is making decisions based on that data.

Data readiness becomes a prerequisite, not an afterthought. If you're migrating into Salesforce, syncing data between platforms, or connecting legacy systems to your CRM, the margin for error just dropped to near zero. That's the kind of work we do at ClonePartner — precision data migrations and custom integrations that hold up when agents, not just people, depend on the data.

What Happens Next

All the updates to Headless 360 are expected to be released in phases. Generally available features include Agentforce Vibes 2.0, the DevOps Center MCP, Session Tracing, and the Agentforce Experience Layer. Features that are in early access include Custom Scoring Evals. Other features, such as the Testing Center and the Salesforce Catalog, are scheduled for rollout in May and June, respectively.

Expect other SaaS vendors to follow. CIO described Salesforce's move as a repositioning from AI agents inside Salesforce toward Salesforce as a programmable platform for agents operating across external tools, interfaces, and environments — a bigger architectural claim than a feature launch. (cio.com)

The other likely outcome is an ecosystem war over context. Analysts told CIO that modern data stacks can replicate much of Headless 360's functionality with more flexibility and less vendor concentration. Vendors like K2view and Spotlight.ai are already pitching MCP and knowledge-graph layers on top of Salesforce because raw tool access doesn't automatically give an agent useful business context. (cio.com)

Salesforce is already shifting from per-seat to consumption-based pricing — a tacit acknowledgment that when agents, not humans, are doing the work, charging per user no longer makes sense.

The developer tooling is available now in the free Salesforce Developer Edition, with usage limits of 110 requests per month with Claude Sonnet 4.5 and 1.5 million tokens per month, refreshing monthly through May 31, 2026.

For buyers, the near-term question isn't whether AI replaces the Salesforce UI tomorrow. It's which workflows can safely move off the browser first. A reasonable starting set: read-only reporting, service context retrieval, and developer automation. Check feature availability carefully — and check pricing even more carefully.

For a deeper look at how Salesforce Service Cloud is evolving into an AI-powered helpdesk, see our complete Salesforce Service Cloud guide.

Salesforce Headless 360 FAQ

What is MCP (Model Context Protocol)?

MCP is an open-source standard, originally created by Anthropic, that gives AI agents a universal way to connect to external tools and data sources. Model Context Protocol (MCP) is an open-source protocol that makes it easier for AI apps and agents to connect with tools and data sources. Before MCP, each tool needed its own custom integration code. With MCP, you build one MCP server for your tool, and it can plug into any AI app or agent that understands MCP. In Salesforce, Hosted MCP Servers expose data, automations, prompts, and analytics through standard OAuth-based access. (docs.anthropic.com)

When was Salesforce Headless 360 launched?

Salesforce confirmed the launch of Salesforce Headless 360 on April 15, 2026, at its TDX developer conference in San Francisco. TDX 2026 ran April 15–16. Features are rolling out in phases through mid-2026.

Is Salesforce Headless 360 free?

The MCP tools and Agentforce Vibes IDE are included in the free Salesforce Developer Edition with usage caps (110 requests per month with Claude Sonnet 4.5, 1.5M tokens per month through May 31, 2026). Industry analysts have raised questions about pricing that Salesforce has not publicly addressed, including whether all tools are included at no cost. Broader enterprise pricing has not been disclosed. (developer.salesforce.com)

Is Headless 360 the same as Agentforce?

No. Agentforce is Salesforce's AI agent product — you use it to build and run agents. Headless 360 is the infrastructure layer that makes the entire platform accessible to any agent, including Agentforce but also third-party tools like Claude Code, Cursor, and OpenAI Codex.

Do I need to be a developer to use Headless 360?

Headless 360 is primarily a developer story for now. Salesforce has positioned it as a way to expand platform access through AI-assisted coding, but the current setup requires an admin or developer who can handle OAuth, external client apps, permissions, and MCP server enablement. (developer.salesforce.com)

---

Salesforce just told its entire user base: you may never need to open the CRM again. Whether that's exciting or unsettling depends on how ready your data and integrations are for the agents that are coming. The winners won't be the teams with the flashiest demos — they'll be the teams with the cleanest data, the clearest permissions, and the best-connected systems.