SAN FRANCISCO, April 20, 2026, 08:55 PDT
- Salesforce rolled out Headless 360, opening up platform features through APIs, MCP tools, and command-line actions tailored for AI agents. Over 60 new MCP tools and upwards of 30 coding skills are on offer from the start.
- Software firms are launching this as they look to address investor concerns over AI’s potential to erode standard software pricing and upend business models.
- Salesforce reported Agentforce ARR hitting $800 million for fiscal 2026. According to Fortune, the company’s own use of AI agents shaved $100 million off its yearly support expenses.
Salesforce rolled out Headless 360, aiming to let AI agents interact with its products without anyone touching the CRM interface. Revealed during the TDX developer event in San Francisco, the suite exposes Salesforce features as APIs, Model Context Protocol resources, and command-line operations for direct use by software agents.
The timing is critical. Earlier this year, software stocks took a beating as investors grew wary that rapid-fire AI advances could threaten the traditional seat-based software business. Salesforce still faces skepticism, with Agentforce yet to demonstrate it can move past limited pilots and into routine use. Back in February, Reuters put the software and services index’s losses at roughly $1 trillion in market value during the downturn, and analysts flagged the need for Salesforce to show real traction with larger enterprise clients.
The competition is heating up. In February, OpenAI introduced Frontier, a tool aimed at helping firms set up and control agents using their current infrastructure. Then on Monday, Adobe joined in with CX Enterprise, its new AI-agent product targeting marketing and customer functions. The wave of launches signals that major software players are rapidly reshaping their platforms for automation.
Salesforce is pitching Headless 360 as a package with three core components: developer tools, a new experience layer able to handle workflows on Slack, Voice, WhatsApp and more, plus governance options for agent testing and optimization both pre- and post-launch. MCP, which Anthropic debuted, acts as an open standard—letting AI agents tap into external tools and data sources directly, instead of requiring human clicks to move things along.
Salesforce announced immediate availability for over 60 new MCP tools and more than 30 ready-made coding skills, offering coding agents like Claude Code, Cursor, Codex, and Windsurf real-time hooks into customer data, workflows, and business logic. VentureBeat reported the full package brings the total above 100 tools and skills. CIO, citing analysts, noted Salesforce wants to position itself as the orchestration layer for agent-driven tasks, extending even beyond its platform.
Salesforce is pushing its case with customer stories. Engine, which runs a business travel platform, announced in a press release that it built Eva—its customer-service agent—on Agentforce in just 12 days. The company now routes half of its chat cases to the software, no humans needed. Support handling time, they say, dropped 15%, while time spent on sales research is down 40%.
Over the weekend, Fortune revealed that Salesforce has been running Agentforce behind the scenes, managing 3 million customer conversations and shaving $100 million off its annualized support tab. “The ‘real unlock’ was capacity as much as cost,” Salesforce president of customer success Jim Roth told the magazine. That idea sits squarely at the core of Salesforce’s updated investor pitch: AI isn’t just a cost-cutter, it’s a revenue driver. Fortune
There’s some backing for that claim. Salesforce reported fiscal 2026 revenue up 10% to $41.5 billion. Agentforce ARR surged 169% year over year, hitting $800 million, with cumulative Agentforce deals now at 29,000. For fiscal 2027, the company projects revenue between $45.8 billion and $46.2 billion, and expects organic growth to pick up in the back half.
Marc Benioff is positioning the effort as a direct response to “SaaS-pocalypse” worries. Back in March, Reuters highlighted how software leaders were pushing back against the notion that AI models might destroy legacy business apps. For Salesforce, the company’s argument partly hinges on its trove of proprietary customer and workflow data—something competitors can’t easily build themselves. Reuters
The gap between Salesforce’s sales pitch and actual results remains. Back in February, Rebecca Wettemann, Valoir CEO, told Reuters the company still has to prove customers are graduating from pilot programs to full-scale deployments. Info-Tech’s Scott Bickley, quoted by CIO, flagged pricing as a sticking point and urged buyers not to get locked in before nailing down the numbers on Headless 360. That risk hangs over the story: Salesforce could expand its footprint, but only if businesses are willing to shoulder the costs, governance headaches, and vendor lock-in baked into broader automation.
