Salesforce announced Headless 360 at TrailblazerDX 2026 on April 15, rebuilding the platform so that every capability is accessible as an API, MCP tool, or CLI command. The release includes more than 60 new MCP tools and 30 preconfigured coding skills that give Claude Code, Cursor, Codex, and Windsurf complete, live access to customer data, workflows, and business logic inside any Salesforce org.
“Why should you ever log into Salesforce again?” Co-Founder Parker Harris asked last month, framing the direction. The answer shipped today: agents call APIs, invoke MCP tools, and run CLI commands directly. No browser required.
What Headless 360 Delivers
The 60+ MCP tools expose Salesforce’s full stack to external coding agents. A developer using Claude Code or Cursor can now read Salesforce data, create records, trigger flows, and execute business logic as part of an agentic workflow without writing custom API integrations for each Salesforce object type. The 30 preconfigured coding skills provide pre-built agent behaviors for common Salesforce development patterns, according to SalesforceBen.
The DevOps Center MCP brings programmatic access into CI/CD pipelines. Natural Language DevOps lets developers describe what they want to deploy and have agents handle the execution. Salesforce claims the integrated build loop cuts cycle times by as much as 40%.
Native React support is also included, letting developers who want full control over the visual layer build custom interfaces with any design language or interaction model.
Agentforce Vibes 2.0: Multi-Model Support Arrives
Alongside Headless 360, Salesforce shipped Agentforce Vibes 2.0, the upgraded version of its vibe-coding tool. The significant change: multi-model support. Developers can now choose Claude Sonnet or GPT-5 as the underlying model, turning Agentforce from a single-model product into a bring-your-own-model platform.
“Since Agentforce Vibes came out just prior to Dreamforce ‘25, this tool has changed the way many work with Salesforce metadata,” wrote Tim Combridge, technical content writer at SalesforceBen. “Now, those of us who prefer Claude Sonnet or another model can connect it to do what we need to do. A huge, welcome change.”
Conversation as the Interface
Salesforce is positioning Slack as the primary surface for agent interaction. Custom AI agents on Slack have grown 300% since January, the company said. The new Agentforce Experience Layer separates what an agent does from how it appears, rendering interactive components like approval workflows, decision tiles, and data layouts natively inside Slack, ChatGPT, Claude, Gemini, Teams, or any MCP-compatible client.
“Not just text back and forth, but approvals, decisions, rich data, full workflows, all surfaced inside the channels where your people already are,” Salesforce stated. “Build once, render everywhere your people already work.”
Agent Fabric, the multivendor agent control plane launched in September 2025, also expanded with automated agent discovery, centralized LLM governance, and deterministic orchestration for companies running agents across multiple platforms and vendors.
Agent Governance: Testing, Observability, A/B Testing
Salesforce addressed the reliability challenge head-on. Testing Center surfaces logic gaps and policy violations before launch. Custom Scoring Evals go beyond pass/fail to score whether an agent made the right decision against developer-defined standards. Agent Script gives explicit control over which parts of an agent’s behavior follow deterministic business logic and which parts can reason freely.
Post-launch, Observability and Session Tracing show why an agent behaved a certain way. A/B Testing runs multiple agent versions against live traffic simultaneously.
One enterprise customer saw Agentforce adoption jump from 22% to 78% in six weeks after the experience layer was updated, according to SiliconANGLE. The agent did not change; the interface did.
Copado Ships Agentia on the Same Day
In a same-day announcement, Copado launched Agentia on April 14, embedding context-aware AI agents directly into the Salesforce DevOps pipeline. While Headless 360 opens the platform to external coding agents via MCP, Agentia works from the inside: agents that understand a team’s specific Salesforce metadata, deployment history, test coverage, and pipeline stages, then participate in the software delivery lifecycle.
“While many vendors are still experimenting with AI features, Agentia brings governed agents directly into production delivery workflows,” CEO Ted Elliott said. Copado described the approach as “AgentOps,” a term for managing and governing intelligent agents within enterprise delivery systems.
Agentia includes specialized lifecycle-aware agents, a primary orchestration agent for multi-step workflows, and Agentia Studio for designing custom agents within Salesforce Agentforce.
The 150,000-Developer Shift
Together, these three announcements represent a single trajectory: Headless 360 opens Salesforce’s data and logic layer to any external coding agent via MCP. Agentforce Vibes 2.0 brings multi-model agent development natively inside Salesforce. Copado’s Agentia embeds governed agents into the deployment pipeline that gets that code to production.
For the 150,000+ Salesforce developers globally, the practical change is immediate. If a customer runs on Salesforce, builders can now construct agents that read, write, and execute directly against that CRM without writing custom API code for every object type. The MCP surface is the bridge. The question that follows is security: 60+ new MCP-callable tools represent a significant attack surface that will require the same authorization discipline the broader MCP ecosystem is still working to standardize.