Salesforce Headless 360, Gemini Legal, and WSO2 ship today

Three launches today push AI agents past the chat window and into governed systems: Salesforce opens trusted business capabilities over MCP, Google Cloud puts a legal-specific agent stack into preview, and WSO2 lets regulated teams self-host the control plane that polices models and MCP traffic.

Salesforce expanded Headless 360 so authorized AI agents can discover and use Salesforce business capabilities through open standards instead of hand-built integrations for every new surface. The centerpiece is the Headless 360 MCP Server, now in open beta. Salesforce says agents in Agentforce, Claude, ChatGPT, Cursor, and other platforms can dynamically discover, understand, and invoke Salesforce capabilities while inheriting existing identity, permissions, workflows, and governance. 1

Alongside it, Salesforce lists the Data 360 MCP Server as generally available, exposing nearly 200 Data 360 APIs so authorized agents can build semantic models, transform data, create segments, and activate campaigns through natural language. Slackbot’s MCP Client is also generally available, connecting Slackbot to Salesforce MCP servers and more than 20 partner apps. Skills for Data 360 are listed as generally available later in August 2026.

The practical signal is capability reuse under existing enterprise controls. Agents stop stalling at the edge of one app only if the permissions and business rules travel with every action. Treat the open beta as a controlled trial on non-destructive workflows, and keep a human review path for anything that changes customer records or money movement.

Google Cloud launched Gemini Enterprise for Legal in preview as a purpose-built industry package on Gemini Enterprise. The system pairs legal skills, MCP connectors into systems firms already use, agents that carry work through a matter, and a partner ecosystem, with a governed control plane underneath. Skills cover contract review and redlining, playbook creation, regulatory horizon scanning, legal research, and DSAR fulfillment. 2

Connectors reach productivity suites, document management systems such as iManage and NetDocuments, Docusign, e-discovery tools, and research sources. Google says access inherits existing role-based controls, document-level permissions, and ethical walls rather than flattening them. Client data and firm playbooks stay private to the organization and are not used to train Google foundation models. A sister preview, Gemini Enterprise for Financial Services, shipped the same day with a Financial Research agent and market-data MCP connectors. 3

This matters because general-purpose AI is not enough for privileged work. The boundary to watch is permission inheritance and citation grounding under real matter walls. Preview means pilot on reversible workflows with attorney review before anything client-facing.

WSO2 made AI Workspace, its AI control plane, fully self-hostable inside the self-managed API Platform. Regulated enterprises and governments can now run governance for LLMs and MCP servers inside their own boundary, including fully air-gapped deployments with no outbound connection to WSO2. 4

Previously, customers could self-host the AI gateway but not the control plane that governed it. The self-managed release closes that gap: employee access to external and sovereign models, exposure of internal resources as MCP servers under the same policies, cost caps, quotas, chargeback, and consistent guardrails across AI gateway traffic. It is generally available via installer, virtual machine, Docker, or Kubernetes.

For platform teams under DORA, NIS2, or the EU AI Act, the useful claim is sovereignty of policy enforcement and audit trails, not a new model. Start with one approved model path and one internal MCP surface, and verify that logs never leave the chosen boundary.

Taken together, today’s launches are about authority: who may act, which systems they may touch, and who keeps the audit trail. The useful next step is not more autonomy. It is a smaller surface you can inspect before production.