Arcade.dev is built for authorization and governance — an agent acting as a specific end-user, with that user's actual permissions, backed by real enterprise funding. AgentRaaS doesn't have that identity model; it's org-scoped, and its whole reason to exist is exactly-once execution. Worth knowing which gap you're actually trying to close.
Arcade's three pillars are authorization, reliability, and governance. AgentRaaS is deliberately narrower: exactly-once execution, proven, not asserted.
| Dimension | AgentRaaS | Arcade.dev |
|---|---|---|
| Core job | Exactly-once execution, dedup, circuit breaker, audit log | Authenticated tool-calling with per-user authorization and governance |
| Identity model | Org-scoped API keys | Per-end-user delegated authorization — the agent acts as a specific person |
| Tool catalog | ~22 curated + self-serve HTTP and MCP Custom Actions | 8000+ MCP tools |
| Self-hosted, free | Yes, unlimited, forever, MIT/Apache-2.0 core | No self-host free tier |
| Entry price | Free tier, $20/month Pro | Free hobby tier (100 auth challenges, 1000 tool executions/mo), $25/month next tier |
| Funding | Solo-built, bootstrapped | $72M total, $60M Series A (June 2026) — SYN Ventures, Morgan Stanley, Wipro |
If your real requirement is "the agent should only be able to do what this specific logged-in user is allowed to do," that's an identity/authorization problem AgentRaaS doesn't solve today — Arcade was built specifically for it, and their funding reflects real enterprise demand for that governance layer. AgentRaaS's org-level API keys are a different, coarser model.
Sources: Arcade Raises $60M — Enterprise DNA, Arcade Review — The AI Agent Index