Product Hunt Finds·Nov 03: MCP Gateways & Agent Teams

There are a few new faces on Product Hunt today. The number one spot goes to AgentKit. I froze when I saw the tagline.

Is that the MCP gateway? I saw it this morning too. Made by Scalekit.

Yeah. It says it lets agents call over thirty thousand tools. Covering six hundred plus apps. Those numbers are kind of scary.

Don't let the numbers intimidate you. That's not the point. The key is the old problem it solves.

What problem? My first reaction was, another tool aggregation platform.

Authorization. Auth is the hardest part of production agents. Every action has to run as the person who initiated it.

Wait. You mean every call carries the user's permissions?

Exactly. And within the scope the user authorized. That's actually a hard compliance requirement.

That really hits a sore spot. When we did our AI product line, permissions were a nightmare.

It also has Virtual MCP Servers. Each agent only gets the tools its role needs.

That's just the principle of least privilege. But applying it to agents is genuinely new.

There are also Triggers. An email lands and it spins up an agent. That turns agents from manual triggers into event-driven.

So the business model should be clear. Ten thousand free calls, then you pay.

Classic infrastructure play. Developers adopt it first, usage grows, and they pay naturally.

So it grabbed the top spot. I'm guessing because it hit the timing of agents going into production.

To sum up. AgentKit isn't selling lots of tools. It's filling the gap of auth and auditing.

Okay. Number two. Teamily AI. I'm more interested in this one.

Pretty bold tagline. Your personal agent, agent teams, and people, all in one messenger.

It emphasizes multi-person, multi-agent collaboration. Not a single-user personal assistant.

So it's up against Slack or Feishu. But it puts agents into group chats as full members.

Agents have persistent memory and can act proactively. Sounds like giving every team a digital coworker.

Sharing context across multiple users and agents. That's actually very hard engineering. Think about how to isolate permissions and memory.

It stuffs docs, meetings, calls, and workflows into one AI-native messenger. That's ambitious.

But there's a contradiction. The broader the features, the harder it is to ensure depth in each.

True. But it got over a hundred and fifty votes and fifty-plus comments. That shows people are really waiting for this direction.

What moves people, I think, is turning agents from tools into coworkers. That mindset shift matters.

Then number three. Testifly. Autonomous AI testing. Zero prompts. Zero maintenance.

That positioning is spot on. CTOs and founders of small teams. No time to write tests, no resources to maintain them.

It claims to discover flows, write tests, self-heal, and even explain failures with video.

Self-healing is key. The most expensive part of traditional automation testing is maintenance. A page change breaks everything.

Hook it up to your web app and it handles it all. Thirty-day free trial. That's a smooth conversion path.

I think it's actually on the same line as AgentKit. Both let agents do real work.

Now that you mention it, the common thread among today's three products is clear. It's all about agents landing in workflows.

Right. From infrastructure to collaboration space to vertical QA. Layer by layer. That shows agents are really moving into production.

So today's recommendation. If you could only try one, who would you pick?

I'd pick AgentKit. You can't avoid the auth and auditing hurdle. It's the foundation. Which would you pick?

I'd rather try Testifly. Zero-maintenance testing is a real relief for small teams. One fills the foundation, one saves manpower.