Arcade.dev

Arcade.dev is an MCP runtime and secure action layer that gives production AI agents user-scoped authorization, reliable tool-calling, and full governance across enterprise systems. It's built for teams moving agents from pilots to thousands of production workflows without over-privileged service accounts.

Arcade.dev is a Model Context Protocol (MCP) runtime platform — 'the secure action layer for enterprise AI agents' — that handles authorization, execution, and governance so AI agents can safely take actions in production. It solves three barriers to deploying agents: authorization, by letting agents act with the exact access a user has for only the action they're taking, with credentials that never leave the runtime and no standing or over-privileged permissions; reliability, via 8,000+ agent-optimized tools purpose-built so user intent maps to the correct API call the first time rather than thin API wrappers that hallucinate; and governance, through a central control plane that logs every action — which agent, on behalf of which user, against which resource. Arcade is framework- and model-agnostic (works with Claude, Codex, LangChain, LlamaIndex, CrewAI, and custom agents), integrates with enterprise identity providers such as Okta, Entra, and SailPoint, offers Python and JavaScript SDKs, and supports cloud, on-premises, air-gapped, and hybrid deployment with SOC 2, SSO, RBAC, and full audit logs. Led by CEO Alex Salazar, Arcade authored the MCP authorization specification adopted by Anthropic, runs in production at major banks, industrial, and pharmaceutical firms, and raised a $60M Series A in June 2026 led by SYN Ventures with Morgan Stanley and Wipro, bringing total funding to $72M.

Category

AI Agents & Orchestration

Pricing

Freemium, Usage-based

Target Market

CTOs, Enterprise Developers, Platform Engineers, CISOs

• ✓
  Agent authorization

  Agents act as authenticated users with dynamic, IdP-based permissions scoped to a single action; credentials never leave the runtime.

• ✓
  Agent-optimized tools

  8,000+ MCP tools purpose-built so user intent maps to the right API call the first time, reducing hallucinations and retries.

• ✓
  Central control plane

  Unified auditing, policy enforcement, and action logging of which agent acted on behalf of which user against which resource.

• ✓
  Framework and model agnostic

  Works with any LLM and agent framework — Claude, Codex, LangChain, LlamaIndex, CrewAI — and any MCP client.

• ✓
  Enterprise identity integration

  Connects to Okta, Entra, and SailPoint with SOC 2, SSO, RBAC, and full audit logs across cloud, on-prem, air-gapped, and hybrid deployments.

• •
  Move agents from pilot to production

  Enterprises deploy agents that act securely across thousands of workflows without over-privileged service accounts.

• •
  Secure tool-calling

  Agents reliably invoke Salesforce, Google Workspace, Slack, and Microsoft actions with the calling user's own permissions.

• •
  Agent governance and audit

  Security teams get a complete trail of every action an agent takes, on whose behalf, against which system.

Free tier available; usage-based pricing designed for enterprise scale.

This page was written from 2 sources.

1. 1.arcade.dev — arcade.devvendor
2. 2.arcade.dev — arcade series avendor