Arcade raised $60 million to solve an overlooked problem blocking enterprise AI agents: control over what they can do. The San Francisco startup focuses on governing AI agents’ actions inside company systems. This is the real bottleneck stopping AI agents from graduating beyond pilot projects.

AI agents don’t fail because their models are flawed. They fail because no one can prove which agent, acting for which user, is allowed to perform a specific action on a given system. Arcade’s CEO Alex Salazar, a former Okta product lead, calls this the “authorization wall.” It’s not identity verification — companies already know who agents claim to be — it’s what they’re allowed to do that matters.

Arcade’s platform acts as a strict gatekeeper. It enforces user-scoped permissions at runtime, so agents inherit only the exact rights of the user they represent. This eliminates the risk of AI agents exploiting overprivileged access or causing unintended damage. The startup supports over 8,000 prebuilt tools designed specifically for AI workflows, far outpacing competitors.

The company’s technology is built on the Model Context Protocol (MCP), an open standard for connecting AI models to real-world tools and APIs. Arcade authored the MCP authorization specification, which was adopted by Anthropic. Its runtime checks every agent action against live permission data, logs every move, and can run inside customer environments — cloud, on-premises, or hybrid.

Arcade’s approach is neutral and vendor-agnostic. It doesn’t rely on any single AI model or platform. This gives companies the flexibility to run multiple models securely under one governance layer. Arcade’s governance plumbing is the foundation on which enterprises can safely build autonomous AI workflows.

Funding and Market Context

Arcade’s $60 million Series A round was led by SYN Ventures, with strategic investors Morgan Stanley and Wipro joining in. Combined with a $12 million seed round raised last year, Arcade has now secured $72 million. SYN Ventures managing partner Jay Leek joined Arcade’s board, emphasizing the urgent need for infrastructure that makes AI agents safe to scale in production.

Tool call volume on Arcade’s platform surged 25-fold in six months, reflecting growing enterprise adoption. Major customers include a top U.S. bank, Prosus, and LangChain, signaling trust from some of the most demanding sectors. These firms are moving AI agents from demos to real production environments, where governance is non-negotiable.

The investment highlights a clear market gap. Most AI agents stall at pilot stage because enterprises lack the ability to prove who authorized each action. Arcade’s platform closes this gap with a robust audit trail showing which agent acted, on whose behalf, and against which resource. This “who did what” accountability is becoming a baseline requirement as AI governance regulations tighten.

Challenges and Opportunities Ahead

Arcade faces a crowded and fast-evolving ecosystem. Big tech players could build competing authorization layers tailored to their own AI platforms. Arcade must scale its sales, expand its tool catalog, and prove ROI to justify its valuation. The startup bets that its early standards leadership and vast tool ecosystem will create a durable moat.

Financial services represent a high-value target where Arcade’s audit and governance features meet strict regulatory demands. Beyond compliance, the platform’s ability to dynamically update AI agent permissions via integration with identity providers reduces attack surfaces and prevents credential abuse.

Arcade’s founders come from Okta and Redis, giving them deep expertise in identity and infrastructure. Their vision positions the company as the “plumbing” layer for the agent era — the silent enforcer that enterprises must have if they want AI agents acting autonomously without risk.

As AI agents move from chatbots to autonomous operators, the question isn’t whether they will act — it’s who controls what they can touch. Arcade stakes its claim to that control.