# Agent-Facing Product

An Agent-Facing Product is a product built so an AI agent can use it directly on a user's behalf, not just a product with a human-friendly UI. In practice that means the value of the product is reachable through an API, an MCP server, a CLI, or a Skill — the same surfaces an agent looks for when a UI isn't an option. Emmanuel Paraskakis developed this framing for his "What Agents Want from Your Product" workshop.

## Why it matters

The core claim behind this term, from the workshop's own framing: "Your users are about to send agents to do their jobs for them. The agent won't click through your UI, sit through your onboarding, or follow your navigation. It will look for an API, an MCP server, a CLI, or a skill. If it can't find one, it moves on." That's the whole problem in one sentence — a product can be excellent for humans and invisible to the agents those same humans start delegating to.

This isn't a hypothetical framing Emmanuel invented in isolation. It echoes what platform leaders are already saying publicly. Aaron Levie (CEO, Box): "You have to be completely comfortable serving up your value proposition as much through agent on or off your platform, as you are your own interface." Dharmesh Shah (Founder & CTO, HubSpot): "Being agentic is not just about agents running on our platform, it's about agents running our platform (being able to operate it). That's how you take AI from being a simple tool to a savvy teammate."

Most products fail this test today for the same reasons: vague naming, unhelpful errors, docs that get skipped, pagination that breaks, and — most simply — no API, no MCP server, no CLI at all. The fix isn't a UI redesign; it's making sure the product's value is reachable without a UI in the first place.

## How to apply it

Emmanuel's audit approach (the walk-in / walk-out structure of the WAW workshop):

1. Watch what an agent actually does when it hits your product cold — most teams have never seen this.
2. Score your product against a 12-point agent-readiness checklist.
3. Write down (or adopt) an API standards file your team can share, so agent-readiness isn't re-litigated per feature.
4. Install a skill that helps you design agent-ready surfaces going forward, instead of retrofitting after the fact.

Deeper dive: [What Agents Want from Your Product](https://maven.com/emmanuel/what-agents-want).

### FAQ

**Is "agent-facing" the same as "AI-powered"?** No. AI-powered usually means you added an AI feature inside your UI. Agent-facing means an external agent, acting for your user, can operate your product without going through that UI at all.

**Do I need a chatbot to be agent-facing?** No — a chatbot is one UI among many. Agent-facing is about whether your underlying capabilities (API, MCP server, CLI, Skill) are reachable and usable by an agent, regardless of whether you also have a chat interface.

**What's the first sign my product isn't agent-facing?** No documented API, no MCP server, and no CLI. If the only way in is a web UI built for clicking, an agent has nowhere to go.

**Does becoming agent-facing mean giving up my UI?** No. It means your value has to work through both — headless (agent-reachable) and through your own interface — at the same time, not one instead of the other.

**How do I know if this applies to my product yet?** If any of your users could plausibly delegate a task involving your product to an AI assistant in the next 12-24 months, it applies now, before that becomes urgent.

Source: Level 250 — https://www.level250.com/glossary/agent-facing-product