As AI agents transition from pilots to production, a $60M breakthrough and a new standard solve the critical accountability gap.
We are entering the Agentic Era, where AI is no longer just answering questions, but taking actions. Yet, 91% of enterprise AI initiatives remain stuck in pilot purgatory. The bottleneck has shifted from how smart a model is, to who governs what it does.
When an autonomous agent acts, who is responsible? Traditional security tools cannot verify which agent acted on behalf of which user, on which system. This accountability gap stops enterprise deployment cold.
Unlike human employees, AI agents do not fear being fired. They will exhaustively exploit every single permission they inherit to achieve their goals. Standard API gateways simply cannot keep up with this relentless drive.
To solve this, Arcade.dev has secured $60 million in Series A funding, bringing its total to $72 million. Led by SYN Ventures, with backing from Morgan Stanley and Wipro, the race is on to build the secure action layer for AI.
Arcade's philosophy is simple: separate the model's reasoning from tool execution. As CEO Alex Salazar says, 'Agents don't fail because the model is wrong. They fail because nobody can prove who is authorized to do what.'
Enter the Model Context Protocol (MCP), an open-source standard created by Anthropic. Think of it as USB-C for AI—a universal menu that securely connects AI assistants to external tools and data.
To make MCP enterprise-ready, Arcade.dev and Anthropic co-developed URL Elicitation. This allows agents to securely request access tokens from users without the LLM ever seeing sensitive credentials.
But asking users for permission every single time is exhausting. That is why the MCP community just stabilized the Enterprise-Managed Authorization (EMA) extension, bringing zero-touch security to the workplace.
With EMA, corporate identity providers like Okta centrally manage access. Employees log in once and instantly inherit secure permissions for all authorized AI tools, with zero extra steps.
Major platforms like Anthropic and Microsoft are already integrating EMA into Claude, Claude Code, and VS Code. Ramp provisioned 2,000 employees instantly, proving that secure autonomy is finally ready for prime time.
While Okta leads the charge, the industry is racing to support other identity systems. The tension between developer agility with local tools and strict corporate compliance will shape the next phase of the Agentic Era.
With secure execution governance, enterprises can finally release AI agents into the wild safely. The foundation is laid, the standards are set, and the era of true autonomous productivity has begun.
Discover more curated stories