What Arcade does
Arcade (Arcade AI, Inc.) is a private enterprise agent-infrastructure company building the “actions runtime” that sits between AI agents and the business systems they need to operate. Arcade’s core premise is that production agent deployments are blocked less by model capability than by secure, multi-user authorization and enforceable governance at the moment an agent tool call would take real action. Arcade therefore provides (1) delegated authentication/authorization anchored to the customer’s identity provider, (2) per-action policy enforcement and least-privilege scoping for user-specific permissions, (3) reliable tool execution with agent-optimized integrations (including MCP tool support), and (4) centralized governance/auditability via a single control plane that exports runtime records for security and compliance workflows.
Product-wise, Arcade is positioned around enterprise deployment of agent tool access through MCP (Model Context Protocol): it supports deploying secure tool execution for agents via an actions runtime that integrates across MCP clients/servers, and it also provides gateway-style distribution of tool access through OAuth/SSO flows so users can use a curated set of tools under their own credentials.
Strategically, Arcade has recently focused on scaling production readiness—e.g., introducing pre-authorization to reduce mid-task consent interruptions, expanding governance features like RBAC, and joining broader ecosystem security/governance initiatives such as NVIDIA’s Open Secure AI Alliance. The company has also moved to broaden its MCP ecosystem footprint via its acquisition of Smithery (a public registry/hosting platform for MCP servers).