Arcade

Arcade is not a GTM system or an AI SDR. It is infrastructure for agents that must securely take actions in systems such as CRM, email, productivity, or developer tools. Arcade’s public material separates its managed production runtime from arcade-mcp, an MIT-licensed framework that developers can use to build custom MCP servers with authentication and tool definitions.

• Per-action agent authorization tied to user permissions and identity providers
• Managed OAuth, token, API-key, and secret handling for agent tools
• MCP tool execution, retries, and error-handling guidance
• Central governance with a shared registry, versioning, visibility filtering, audit logs, and OpenTelemetry export
• Prebuilt integrations across productivity, developer, search, sales, and other systems
• Open-source arcade-mcp framework for custom MCP servers, including local and offline use

Engineering, platform, security, or GTM-automation teams building agents that need to take actions in systems on behalf of real users and must satisfy authentication, permissioning, audit, and deployment requirements. This is our interpretation of the authorization and governance focus in Arcade’s published material.

Arcade is a poor fit for a team that only needs a single local read-only script or a one-off internal prompt and has no user-authorized actions to govern: its central value is managing OAuth, user permissions, tool authorization, and auditability. It is also a poor fit for a buyer who wants a sales-engagement product out of the box; Arcade can connect sales tools, but it is the runtime and tool layer rather than the prospecting, sequencing, or CRM application itself.

Setup assessment: custom

A local custom MCP server can be started from the framework in a few commands, but production use depends on the target tools, OAuth configuration, user authorization, identity provider, security policies, and deployment model. Arcade advertises cloud, VPC, on-premises, air-gapped, and hybrid approaches for that reason.

Prerequisites

• An agent use case that needs external actions
• Owners for identity, authorization, and security policy
• Credentials and OAuth configuration for target systems
• A chosen deployment model

Machine-readable API

No OpenAPI link in the reviewed source set

• A machine-readable OpenAPI specification was not included in the reviewed source set.