Blocks.ai Review (2026): Pricing, Features & Honest Verdict

TLDR: Blocks.ai is a control plane and network layer for AI agents from PubNub. It lets you connect, control, and monetize agents across any framework or provider without opening ports, setting up tunnels, or touching your firewall. Best for developers building agents who want them to reach and call each other. Price: free to connect, pay per call with a 15 percent platform cut. Rating: 7.0/10.

We found Blocks.ai on Product Hunt after PubNub launched it in July 2026, and it is one of the more ambitious infrastructure plays we have seen in the agent space. Most agent tooling right now is about building agents. Blocks.ai is about connecting them: making an agent reachable from anywhere, letting it call other agents, and handling the billing when someone pays to use it. It is plumbing, but it is plumbing that the current wave of agent development actually needs.

What Is Blocks.ai?

Blocks.ai is a global network that connects and controls AI agents across frameworks, providers, and APIs. The core problem it solves is reachability. Right now, exposing an agent to the outside world means dealing with ports, tunnels, DNS, and firewall rules, and none of that is fun. Blocks.ai makes an agent reachable without any of it. The same connection that makes your agent callable also lets it call other agents on the network, regardless of what framework they were built with.

It comes from PubNub, which has spent years building real-time messaging infrastructure at scale, so the networking pedigree is real. The network ships with a catalog of discoverable capabilities that agents can call, things like code review, transcription, PDF and QR generation, data extraction, and streaming compute. The vision is an interoperable web of agents that can find and use each other, with Blocks.ai handling the security, streaming, billing, and observability underneath. It is early, but the direction is clear.

What Are Blocks.ai’s Key Features?

Zero-config agent reachability

This is the feature that will make developers pay attention. You connect an agent to Blocks.ai and it becomes reachable, with no inbound ports opened, no tunnels stood up, no DNS changes, and no firewall edits. Anyone who has fought with ngrok tunnels or reverse proxies to expose a local service knows how much pain this removes. Your agent connects outbound to the network and becomes callable, which is both simpler and safer than punching holes in your infrastructure.

Agent-to-agent calling

Once an agent is on the network, it can call other agents on the network, even ones built with completely different frameworks. This is the interoperability piece, and it is what separates Blocks.ai from a plain hosting layer. An agent that only does one thing can reach out to a transcription agent or a code-review agent and compose capabilities it does not have itself. Whether this becomes genuinely useful depends entirely on the catalog filling up, but the mechanism is sound.

Twenty production capabilities included

Blocks.ai bundles the boring but essential infrastructure that every serious deployment needs: task queuing and load balancing, enterprise authentication with JWT and OIDC, real-time bidirectional streaming, observability dashboards, multi-tenant org management, an agent discovery registry, horizontal auto-scaling, artifact and file handling, webhooks, and audit trails. These are the things teams usually build themselves over months. Getting them out of the box is a real head start.

Built-in monetization

If you build an agent, you can set a per-task or per-minute price and let others pay to call it. Stripe processes the payments, Blocks takes 15 percent, and you keep 85 percent. This turns an agent from a personal tool into something you can list in a catalog and earn from. It is an app-store-style model for agents, and it is the part that could make Blocks.ai more than just infrastructure if the marketplace actually develops.

How Much Does Blocks.ai Cost?

Blocks.ai does not use traditional tiered pricing. It is free to connect and free to list your agents, and callers can try public agents in the browser without paying. When a paid agent gets called, the builder sets the price, Stripe handles the transaction, and Blocks keeps 15 percent while the builder keeps 85 percent. So a $10 call sends $8.50 to the builder and $1.50 to Blocks. There is also Blocks Enterprise, a private instance of the network with dedicated infrastructure and custom commercial terms.

The connect-free, pay-per-call model is fair and lowers the barrier to trying it, since you owe nothing until real usage happens. The 15 percent cut is in the normal range for a marketplace, sitting below Apple’s 30 percent and in line with many creator platforms. The catch for anyone running high volume is that a percentage cut has no ceiling, so at scale a flat infrastructure bill might be cheaper. And because costs are usage-based, they are harder to predict up front than a fixed monthly plan.

Who Is Blocks.ai Best For?

Use Blocks.ai if you are building AI agents and you want them reachable and callable without wrestling with networking. It is especially compelling if you want to monetize an agent, because the billing and marketplace mechanics are already built. Teams that need production infrastructure like auth, observability, and auto-scaling without building it themselves will get real value here too.

Skip Blocks.ai if you are just prototyping a single agent that never needs to talk to the outside world, or if you are not sold on agent-to-agent networking as a concept yet. This is infrastructure for a future that is still forming. If your agents live entirely inside one app and one framework, you do not need a network layer. The value scales with how much you want agents to interoperate and earn.

What Are the Best Blocks.ai Alternatives?

LangGraph Platform

LangChain’s deployment layer for agents, with hosting, persistence, and monitoring. It is more opinionated and tied to the LangChain ecosystem, and it focuses on running your agents rather than networking them together across frameworks. If you already live in LangChain, it is the natural path. Blocks.ai is more framework-agnostic and more focused on connectivity and monetization.

Cloudflare Workers and Tunnels

The DIY route. Cloudflare gives you tunnels, edge compute, and networking primitives you can assemble into agent hosting yourself. It is powerful and cheap, but you build the agent-specific pieces like discovery, billing, and agent-to-agent calling on your own. Blocks.ai trades some flexibility for having all of that ready out of the box.

Self-hosting with a reverse proxy

The traditional approach: run your agent on a server, put it behind a reverse proxy or ngrok, and handle auth and billing yourself. It is the most control and the most work. For a single agent it is fine. For a fleet of interoperating, monetized agents, the maintenance burden is exactly what Blocks.ai is trying to erase.

Final Verdict: Is Blocks.ai Worth It?

Blocks.ai is a serious piece of infrastructure aimed at a problem that is about to get much bigger. Making agents reachable without networking pain, letting them call each other across frameworks, and building in monetization from day one is a genuinely good set of ideas, and coming from PubNub, the underlying network is not vaporware. The connect-free, pay-per-call pricing means you can try it with zero commitment, which is the right way to launch something this new.

The honest caveat is that this is early. Agent-to-agent networking is an unproven category, and the payoff depends on the catalog and the marketplace reaching critical mass, which has not happened yet. High-volume builders should also do the math on the 15 percent cut versus flat infrastructure. But if you are building agents and thinking about how they connect and earn, Blocks.ai is worth wiring up now while it is free to start. Rating: 7.0/10.