Honest marketing · What we actually build
We build both. We’ll always tell you which one you’re getting.
In 2026, the word “agent” means roughly one thing in marketing: an LLM is involved somewhere. That’s it. The moment a product touches a language model, it gets sold as an agent — because “agent” sells, and “retrieval pipeline” doesn’t. We think you deserve the distinction, because the two behave very differently once they’re running on your account.
Most of what’s sold as an “AI agent” — for Instagram DMs, for support, for lead capture — is a RAG pipeline: retrieve some context, put it in a prompt, generate a reply, send it. Fixed path. Same order every time. No decisions. It works, and often it’s exactly what you want. It just isn’t an agent.
A block button doesn’t make it an agent, any more than a thermostat is an agent for turning on the heat.
This isn’t pedantry. The two have genuinely different properties — and buyers get burned by the gap.
For most jobs — auto-reply, FAQ, lead capture, moderation — the pipeline is the right choice, and the “agent” branding is overkill you don’t want. You don’t want an autonomous loop with tool access deciding, on its own, to block a customer or send forty DMs. Predictability is the feature.
Half the time, the honest product is worse for being a real agent.
if decide that, never the model. The dangerous button is kept out of the LLM’s hands. That’s a feature.So here’s the promise: we’ll build you a pipeline when a pipeline is right, and an agent when the job needs one — and we’ll tell you, in plain words, which one you have.
Every agent and pipeline we run, in one place — labelled for what it actually is.
Enter the Agents Console