Salesforce Headless 360: CRM Without the Chrome
⏱ 7 min read
Salesforce Headless 360: Discover how Salesforce is evolving toward an API-first, AI-native platform architecture built for the Agentic Enterprise. Whether you want to understand what Salesforce Headless 360 is, how Salesforce MCP works, or how to connect AI agents to Salesforce, this guide covers APIs, Agentforce, Data 360, MCP tools, Slack-native workflows, security, and autonomous AI workflows – without requiring every task to run through the traditional Salesforce interface.
100+
Reusable Skills highlighted in Salesforce’s Headless 360 expansion
API-first
Capabilities accessible to authorized applications and agents
Governed
Access controlled through identity and permissions
Multi-agent
Connections across supported AI clients and workflows
Contents
hide
What Is Salesforce Headless 360? An Introduction
For years, businesses interacted with Salesforce through dashboards, consoles, forms, and manual workflows. Teams logged into Salesforce, navigated interfaces, updated records, processed approvals, and completed tasks one screen at a time.
That model is evolving. Understanding what Salesforce Headless 360 is is becoming important for technology leaders exploring how AI agents can work with enterprise systems.
Salesforce Headless 360 is Salesforce’s approach to making platform capabilities available through APIs, MCP tools, and command-line interfaces so authorized agents and applications can discover and use them. Rather than making the standard user interface the only way to access a capability, this approach allows supported workflows to run through other interfaces and AI clients.
If you are exploring how Agentforce fits into your Salesforce roadmap, Headless 360 provides an evolving set of capabilities for connecting agents with Salesforce workflows, data, and business logic.
What makes Salesforce Headless 360 different?
- Salesforce capabilities can be accessed through APIs and supported MCP tools.
- Authorized AI agents can use tools without navigating the standard Salesforce UI for every action.
- Developers can connect compatible AI coding clients to supported Salesforce development workflows.
- Agent experiences can be delivered through supported interfaces and channels.
- Existing identity, permission, and governance controls remain important when agents access enterprise systems.
What Is Salesforce MCP?
Salesforce MCP lets AI agents and coding assistants such as Claude, Cursor, and ChatGPT connect to Salesforce through the Model Context Protocol. The Salesforce MCP server exposes tools to query data, deploy metadata, and run tests, while Salesforce permissions control what each agent can see and change.
The Model Context Protocol (MCP) is an open standard for connecting AI applications to tools and external context. Instead of requiring every AI client to implement a separate custom integration for each capability, an MCP server exposes supported functions through a common protocol.
Salesforce offers different MCP-related options. The Salesforce DX MCP server, installed with @salesforce/mcp, supports development workflows. The Headless 360 MCP Server is a separate hosted service that exposes Salesforce capabilities through a smaller set of tools backed by a growing library of operations. Check the official documentation for current availability and feature requirements before choosing a server.
Two Ways Salesforce Uses MCP
1. Developer MCP Server
The Salesforce DX MCP server, @salesforce/mcp, connects compatible coding clients to authorized Salesforce orgs. Depending on enabled toolsets, it can support metadata, data, testing, user-management, and DevOps-related workflows.
2. Headless 360 and Agentforce
Salesforce’s Headless 360 approach exposes supported Salesforce capabilities for use by compatible AI agents and applications. Agentforce and external AI clients can participate in connected workflows where the relevant integrations and features are supported and enabled.
Availability differs by server and capability. The Salesforce DX MCP server distinguishes generally available tools from non-GA tools. Salesforce’s official documentation labels the Headless 360 MCP Server as Beta. Do not assume that every MCP capability is generally available in every org or region.
For a broader explanation of protocol choices, see our guide to MCP vs A2A.
Your First Salesforce MCP Use Case
Start with a low-risk, read-only workflow. Connect a coding assistant to a sandbox, allow it to inspect selected metadata, and ask it to explain how an object, flow, or Apex component works. Once the workflow is reliable, introduce testing and deployment tools in a development or scratch org.
- Start read-only: inspect metadata or query a limited set of sandbox records.
- Add testing: enable appropriate test tools and validate the agent’s recommendations.
- Test deployment: use a scratch org or sandbox before considering production changes.
- Expand deliberately: enable additional tools only when the use case requires them.
Salesforce MCP Security Checklist
- Use a dedicated, least-privilege Salesforce integration user or identity appropriate to the architecture.
- Authorize only the orgs the MCP client needs to access.
- Enable only the required toolsets and individual tools.
- Keep production write access restricted and require approval for high-impact actions.
- Review Salesforce sharing, object permissions, and field-level security.
- Monitor tool activity and retain appropriate audit records.
- Validate changes in a sandbox before production deployment.
MCP does not automatically make an agent safe. Access controls, the identity used by the client, the enabled tools, and the safeguards configured in the AI client all influence the actions it can perform.
Kizzy Consulting helps organizations plan Salesforce and AI integrations. For implementation support, explore our AI integration services.
The Rise of the Agentic Enterprise: Why Headless 360 Matters
Salesforce Headless 360 supports the shift toward the Agentic Enterprise – a model in which AI agents can use authorized tools to work across business systems and collaborate with people. Instead of relying exclusively on humans to navigate interfaces, organizations can design workflows in which agents retrieve context, invoke approved actions, and return outcomes through supported channels.
The difference is not that humans disappear from enterprise workflows. Rather, teams can decide which tasks are suitable for automation, which actions need approval, and where human judgment remains essential.
Traditional CRM Workflows
- People navigate interfaces to complete tasks.
- Users manually move information between systems.
- Approvals and record updates may involve several steps.
- Teams often switch between separate tools.
- Automation depends on configured business rules and workflows.
Headless and Agentic Workflows
- Authorized agents invoke supported tools and APIs.
- Workflows can coordinate multiple systems.
- Approval steps can be built into agent processes.
- Teams can interact through supported channels.
- Monitoring and access controls remain part of the design.
The Three Major Innovations in Salesforce Headless 360
1. MCP Tools and Coding Skills for AI Agents
Salesforce’s developer MCP tooling connects compatible AI coding clients to supported Salesforce capabilities. The Salesforce DX MCP server provides configurable toolsets for orgs, metadata, data, testing, users, and other development tasks. The exact tools available depend on the installed version and configuration.
Compatible clients documented by Salesforce include:
- Claude Code – an agentic coding environment.
- Cursor – an AI-assisted code editor.
- Windsurf – an AI coding environment.
- Other compatible MCP clients configured according to their respective documentation.
For the Salesforce-specific setup and configuration steps, see the Salesforce development guide and the MCP vs A2A explained article for protocol background.
2. Agentforce Experience Layer
The Agentforce Experience Layer addresses a practical question: how can an agent’s output reach users when the standard Salesforce interface is not the interaction surface? The answer depends on the supported channel and implementation. Agent logic and the user experience can be designed separately, allowing an approved workflow to present information or request decisions through suitable interfaces.
| Capability | How It Works | Potential Business Value |
|---|---|---|
| Slack-native workflows | Surface supported updates and approval requests in Slack. | Keep decisions closer to where teams collaborate. |
| Cross-channel experiences | Deliver agent results through supported apps and interfaces. | Reuse business logic across suitable experiences. |
| Interactive workflows | Present decisions, forms, and approval steps through supported components. | Make review and escalation more convenient. |
| Reusable architecture | Separate workflow logic from the presentation layer where appropriate. | Reduce duplicated implementation effort. |
3. Enterprise AI Governance and Observability
Autonomous actions require clear controls. Organizations need to know which identity an agent uses, which tools it can invoke, what data it can access, and when a person must approve an action. Salesforce’s available monitoring and testing features vary by product and release, so verify the tools applicable to your implementation.
Testing
Validate agent behavior, edge cases, and policy requirements before launch.
Tracing and Audit
Review available records of tool activity and agent execution.
Evaluation
Measure quality and business outcomes against defined success criteria.
How Does Salesforce Headless 360 Work? The Architecture Explained
Understanding API-First Enterprise Software
Headless architecture separates a system’s business logic and data from the interface used to interact with it. In the Salesforce context, APIs, MCP servers, and command-line tools provide different ways to access supported capabilities. The standard Salesforce interface remains useful, but it is not necessarily the only interaction surface for every workflow.
Think of it as giving authorized applications and agents structured access to selected enterprise capabilities rather than requiring every interaction to pass through the same interface. The available actions still depend on the API or tool, the user’s permissions, and the configuration of the Salesforce org.
Key Components: MCP Tools, Agentforce, and Data 360 Deep Dive
1. Model Context Protocol (MCP) Tools – The Developer Gateway
MCP is an open standard that lets AI clients connect to tools and context through a common protocol. The Salesforce DX MCP server, distributed as @salesforce/mcp, offers configurable toolsets for Salesforce development and administration workflows. For the complete list of supported tools, GA status, configuration flags, and client examples, consult the official Salesforce MCP repository.
For example, a compatible client configuration can use the following arguments:
npx -y @salesforce/mcp --orgs DEFAULT_TARGET_ORG --toolsets metadata,data,testingThis is an illustrative command for a local setup, not a complete universal installation process. First authorize the intended Salesforce org through the Salesforce CLI, then configure the command in the MCP client’s documented configuration file. Use only the toolsets required by your use case. Follow the official repository’s instructions for the current package version and client.
After setup, an AI coding client can use the enabled tools to inspect metadata, query permitted data, run tests, or perform other supported tasks. Deployment and write operations should be tested in a sandbox before production use.
| Toolset | Typical Purpose | Example Use |
|---|---|---|
| Metadata | Retrieve and deploy supported Salesforce metadata. | Inspect an object or prepare a sandbox change. |
| Data | Perform supported data operations within authorized access. | Query selected records in a development org. |
| Testing | Run supported tests and inspect results. | Validate changes before deployment. |
| Org and users | Inspect or manage supported org and user resources. | Review authorized orgs or permission assignments. |
For the full setup walkthrough, security guidance, and examples, read What Is Salesforce MCP? above.
2. Agentforce – AI Agent Orchestration
Agentforce is Salesforce’s platform for building and operating AI agents. Depending on the implementation, agents can use business context, configured actions, instructions, and connected systems to support tasks across sales, service, and operations.
- Defined agent behavior: instructions and supported logic guide how an agent handles tasks.
- Action execution: agents can invoke configured actions and integrations.
- Escalation: workflows can direct sensitive or uncertain decisions to a person.
- Evaluation: teams can test performance against defined criteria and business goals.
- Experience delivery: outputs can be surfaced through supported channels and applications.
Learn more in our guide to what Agentforce is.
3. Data 360 – The Intelligence Foundation
Data 360 helps organizations bring together and use data from Salesforce and connected sources. For AI agents, the quality and relevance of available context influence whether an agent can return useful answers or take appropriate action.
- Unified customer context: bring relevant customer information together from supported sources.
- Data integration: connect supported systems and data sources according to business needs.
- Data quality: address inconsistent, duplicated, or incomplete information.
- Identity resolution: unify records where the implementation and use case require it.
- AI grounding: make suitable information available to agents under the applicable access controls.
Data 360 is not a substitute for data governance. Organizations still need to define ownership, access rules, quality standards, and appropriate use of customer information. Explore our guide to Salesforce Data 360 implementation.
API-First Integration Patterns: How to Connect Salesforce Headless 360
Three Access Methods – Choosing the Right Approach
1. REST and SOAP APIs
Best for: Custom integrations, middleware, and applications that need defined endpoints and predictable request-response behavior.
- Structured API requests and responses.
- Authentication and access controls.
- Fine-grained integration design.
2. MCP Tools
Best for: Connecting compatible AI clients to supported tools and workflows.
- Common protocol for tool access.
- Configurable toolsets.
- Access constrained by the chosen server and identity.
3. Salesforce CLI
Best for: Developer workflows, scripting, testing, and deployment automation.
- Command-line operations.
- Automation and repeatable tasks.
- Integration into development pipelines.
These methods can complement one another. The right design depends on the required workflow, the client’s capabilities, security requirements, and which Salesforce features are available in the target environment. See our Salesforce integration best practices for additional guidance.
AI Agents and Autonomous Workflows: Use Cases in Production
AI agents can support business workflows by gathering information, using approved tools, following defined rules, and returning results. Whether a workflow is suitable for full automation depends on its risk, data quality, integrations, and need for human judgment.
- Lead qualification: summarize lead information, apply approved qualification criteria, and route leads to the right team.
- Contract processing: extract relevant terms, identify exceptions, and route documents for human review.
- Customer service: summarize cases, retrieve relevant information, draft responses, and escalate complex requests.
- Developer assistance: inspect metadata, propose changes, run tests, and support deployment workflows.
- Operations support: gather information from connected systems and prepare recommended next steps.
Agent Governance Framework – Keeping AI Accountable
| Governance Area | Purpose | Benefit |
|---|---|---|
| Testing | Validate agent behavior and exceptions before release. | Identify problems before they affect users. |
| Access control | Limit which records, fields, and actions are available. | Reduce inappropriate access and actions. |
| Monitoring and audit | Review available activity records and operational outcomes. | Support investigation and accountability. |
| Evaluation | Measure output quality and business results. | Make improvement decisions based on evidence. |
Slack-Native Operations: Bringing AI Into the Flow of Work
Slack can serve as a collaboration surface for supported Salesforce and AI workflows. Depending on the integration, teams may receive updates, review recommendations, and handle approval requests without switching to a separate interface for every step.
Build Reusable Workflows for Appropriate Channels
Consider a discount approval workflow. An agent could retrieve the relevant opportunity context, apply configured business rules, prepare an approval request, and send it to an authorized reviewer. The workflow’s delivery channel and interactive features depend on the systems and integrations enabled.
Slack
Share updates and approval requests in team conversations through supported integrations.
Mobile and Web
Deliver suitable agent experiences through connected applications.
AI Clients
Connect compatible AI clients to supported Salesforce capabilities.
Key Benefits of Salesforce Headless 360 for Enterprise Organizations
The potential value of a headless, agent-enabled architecture depends on the use case and how well it is implemented. Organizations should establish a baseline, choose a measurable outcome, and compare results after deployment rather than assume a particular return in advance.
AI-Powered Automation
- Automate suitable repetitive tasks.
- Support consistent execution of defined workflows.
- Provide assistance outside traditional working hours where configured.
- Reduce unnecessary manual handoffs.
Developer Productivity
- Use AI assistance to inspect code and metadata.
- Support testing and repeatable development tasks.
- Reduce context switching for supported workflows.
- Improve documentation and troubleshooting.
Enterprise Governance
- Apply access controls to agent actions.
- Introduce approval gates for sensitive operations.
- Monitor available execution and audit records.
- Test behavior before expanding to production.
Cost and Operational Efficiency
- Identify repetitive work suitable for automation.
- Evaluate integration and maintenance costs.
- Measure cycle time and quality improvements.
- Compare consumption, licensing, and implementation costs.
The Four Systems Powering Salesforce Headless 360
| Layer | Platform or Capability | Purpose | Learn More |
|---|---|---|---|
| System of Context | Data 360 | Bring relevant enterprise data together for supported use cases. | Data 360 guide |
| System of Work | Customer 360 apps and workflows | Business processes, approvals, records, and operational logic. | Integration guide |
| System of Agency | Agentforce and connected agents | Agent orchestration, configured actions, and task execution. | Agentforce guide |
| System of Engagement | Slack and supported experiences | Human-AI collaboration and delivery of workflow results. | Salesforce platform |
Frequently Asked Questions
What is Salesforce MCP?
Salesforce MCP is Salesforce’s support for the Model Context Protocol, which lets AI agents and coding assistants such as Claude, Cursor and ChatGPT connect to Salesforce orgs and use tools to query data, deploy metadata and run tests within Salesforce permissions.
How do I set up the Salesforce MCP server?
Install the Salesforce DX MCP server with npx, authorize the orgs it may use, choose toolsets such as metadata, data and testing, and add the server to an MCP-compatible client like Claude Code or Cursor. Start in a sandbox.
Is Salesforce MCP secure?
It runs with the permissions of the authorized Salesforce user, so field-level security and sharing rules apply. Use a dedicated integration user, enable only the toolsets you need and keep production write access behind approval.
What is Salesforce Headless 360?
Salesforce Headless 360 is Salesforce’s approach of exposing CRM data, Agentforce and Data 360 through APIs and MCP tools, so AI agents and other applications can use Salesforce without its standard user interface.
What is the difference between MCP and APIs in Salesforce?
Salesforce APIs are fixed endpoints that developers code against. MCP wraps capabilities as tools that an AI agent can discover and call by itself, which makes it faster to connect agents to Salesforce.
Conclusion: Salesforce Headless 360 Is Changing Enterprise Software
Salesforce Headless 360 reflects a shift in how enterprise capabilities can be accessed and used. APIs, MCP tools, command-line workflows, and connected AI agents create additional ways to interact with Salesforce beyond the standard user interface.
The best implementation is not necessarily the one that automates the most actions. It is the one that connects the right tools, data, permissions, and business processes to a measurable use case – with testing and human oversight where they are needed.
- MCP connections: allow compatible AI clients to use supported Salesforce tools.
- Agentforce: supports the design and operation of Salesforce AI agents.
- Data 360: helps make connected data available for appropriate use cases.
- APIs and CLI: remain important for integrations, development, and automation.
- Security and governance: help keep agent activity within defined permissions and business controls.
At Kizzy Consulting, we help organizations modernize Salesforce ecosystems for AI-assisted operations, integrations, and workflow automation. Whether you are exploring Salesforce MCP or planning an Agentforce implementation, the first step is to identify a suitable use case and validate the technical and governance requirements.
Ready to Connect AI With Your Salesforce Ecosystem?
Kizzy Consulting helps organizations plan Salesforce integrations, MCP connections, Agentforce workflows, Data 360 initiatives, and AI-powered automation with appropriate security and governance.
Tell us about your Salesforce and AI integration goals:
