What Arcade does
Arcade (arcade.dev) is an agent “actions runtime” and MCP-focused infrastructure layer that enables AI agents to take secure, user-authorized actions in real business systems. The core problem Arcade targets is the gap between agent demos and enterprise production: in production, organizations need verifiable authorization, least-privilege enforcement, and auditability for every tool call executed on behalf of a specific user.
Arcade positions itself as a centralized control plane between AI agents and external tools/systems, brokering OAuth/IDP-backed delegated access and enforcing policy at runtime before a tool executes. The platform is designed for multi-user deployments (not service-account style access), with features intended to keep credentials isolated from LLMs and to ensure that permissions are evaluated at the moment of action. Arcade also extends beyond authorization into governance and operational concerns such as reliable tool execution, audit trails, and administration controls.
Business model: Arcade is sold as a usage-based platform with tiered plans (Free, Team, Enterprise) that include per-authentication events and per-tool-call usage on managed deployments, plus enterprise options such as SSO/RBAC/audit logs and flexible deployment (cloud, VPC, or air-gapped).