We Just Won "Overall Authentication Solution of the Year" — Here's Why It Matters for AI Builders

We Just Won "Overall Authentication Solution of the Year" — Here's Why It Matters for AI Builders

Shawnee Foster's avatar
Shawnee Foster
JUNE 25, 2025
3 MIN READ
COMPANY NEWS
Rays decoration image
Ghost Icon

Arcade.dev just took home "Overall Authentication Solution of the Year" in the 8th Annual AI Breakthrough Awards. And before you roll your eyes at another tech award announcement, let me explain why this actually matters for anyone building AI agents that need to do real work.

The Problem We All Keep Hitting

You know that moment when your perfectly crafted AI agent suggests "I'll schedule that meeting for you" — and then... doesn't? Because it can't?

Yeah, that's the wall everyone's been hitting. Your AI can write beautiful SQL queries but can't actually query your database. It can draft the perfect email but can't send it. It can plan your entire day but can't touch your calendar.

The disconnect: AI agents today are like brilliant consultants with no hands. They can tell you exactly what to do, but they can't actually do it.

What We Actually Built (No Fluff)

Here's what Arcade.dev does in plain terms:

  • Secure OAuth flows that let AI agents authenticate as actual users (not just bot tokens)
  • Pre-built connectors to Gmail, Slack, Notion, Salesforce, and 50+ other tools
  • 7-minute setup from zero to authenticated agent (we timed it)
  • Tool evaluation framework so you can test agent actions before they go live

Think of it as the authentication layer that should have existed from day one of the AI revolution.

Why "Authentication Solution" Actually Means Something

When the AI Breakthrough judges evaluated 5,000+ nominations from 20 countries, they weren't looking for the shiniest demo. They were looking for infrastructure that makes AI useful in production.

Here's what they saw:

# Without Arcade: AI suggests actions
response = llm.chat("Schedule a meeting with the team")
# Output: "I'd be happy to help you schedule that meeting..."
# Reality: Nothing happens

# With Arcade: AI takes actions

manager = ToolManager(api_key=arcade_api_key)

google_tools = manager.get_tools(toolkits=["Google"])
google_agent = create_react_agent(
    model="openai:gpt-4o",
    tools=google_tools,
    prompt="You are a helpful assistant that can assist with"
           " managing a Google account, contacts, and inbox.",
    name="google_agent"
)

# Agent can now actually create the meeting
response = google_agent.invoke([{
    "role": "user", 
    "content": "Schedule a meeting with the team"
}])
# Reality: Meeting appears on everyone's calendar

The Technical Bits That Matter

OAuth 2.0 Done Right: We handle the entire auth flow, token refresh, and permission scoping. Your agent gets exactly the access it needs — nothing more, nothing less.

Tool Evaluation Framework: Before your agent goes wild in production, you can define test cases and use our critics system to verify behavior:

@tool_eval()
def hello_eval_suite() -> EvalSuite:
    """Create an evaluation suite for the Slack tool."""
    suite = EvalSuite(
        name="Slack evaluation",
        system_message="You are a helpful assistant with access to Slack.",
        catalog=catalog,
        rubric=rubric,
    )
 
    # Example evaluation case
    suite.add_case(
        name="Send DM on Slack",
        user_message="Send a DM to Alex to say we won the prize!",
        expected_tool_calls=[
            ExpectedToolCall(
                func=send_dm_to_user,
                args={
                    "name": "Alex",
                    "message": "We won!!"
                },
            )
        ],
        critics=[
            SimilarityCritic(critic_field="name", weight=0.5),
            SimilarityCritic(critic_field="message", weight=0.5),
        ],
    )
 
    return suite

Then run:

arcade evals .

Deployment Flexibility: Run in our cloud, your VPC, or fully on-premises. Because we know you have that one customer who insists on air-gapped deployments.

What This Means for AI Development

Less than 30% of AI projects make it to production. Want to guess why?

It's not because the models aren't smart enough. It's because they can't authenticate, can't access real systems, and can't take actual actions.

This award isn't just recognition — it's validation that the industry is finally focusing on the unglamorous but critical infrastructure that makes AI agents actually useful.

Try It Yourself

Don't take our word (or AI Breakthrough's) for it. Grab an API key and see what happens when your agents can finally touch the real world:

pip install arcade-ai
arcade login
arcade chat

Ask it to check your email, create a calendar event, or post to Slack. Watch it actually do it.

Then maybe you'll understand why we're so excited about building the connective tissue that makes AI agents real.


Want to see what agents can accomplish with more than just chat? Check out Chat.Arcade.Dev→

P.S. — To the 5,000+ companies who entered these awards: respect. Building in AI right now is like changing the engine while the plane is flying. We're all in this together.

SHARE THIS POST

RECENT ARTICLES

THOUGHT LEADERSHIP

The Agent Hierarchy of Needs: Why Your AI Can't Actually Do Anything (Yet)

Your AI can summarize documents you feed it, answer questions about your uploaded PDFs, and explain concepts from its training data. But ask it to pull your actual Q4 revenue from NetSuite, check real customer satisfaction scores, or update a deal in Salesforce? Suddenly it's just guessing—or worse, hallucinating numbers that sound plausible but aren't your data. This disconnect between AI's intelligence and its ability to access real data and take action is why less than 30% of AI projects hav

Rays decoration image
TUTORIALS

How to choose the best Agentic Framework, Part 2: Agentic Delegation

In the previous post in this series, we explored Human-in-the-Loop. Here, we’re exploring Handoffs, which I prefer to call “Agentic Delegation” This post is a companion to a video, I encourage you to watch it! Here’s the experiment setup I’m using the same agentic system. I implemented the same system using three different Frameworks: * LangGraph * OpenAI’s Agents SDK * Google’s Agent Development Kit (ADK) In all cases, the agent uses a “supervisor” architecture, where a single agent re

Rays decoration image
THOUGHT LEADERSHIP

The Day an AI Agent Merged Malicious Code (And What We Learned)

Yesterday started like any other day. Coffee, standup, code review. Then came the call that made me put down everything. A customer's AI agent had been compromised. Not through some exotic zero-day or sophisticated attack vector. No, this was far more elegant—and terrifying. Their LLM-powered browser agent had autonomously merged a malicious pull request on GitHub. As a real employee. With real permissions. The attack vector? A carefully crafted email sitting in the user's inbox, containing in

Blog CTA Icon

Get early access to Arcade, and start building now.