---
schema: 1
type: term
title: '**agent-native**'
description: >-
  **agent-native** — describing software designed so AI agents can act within it
  — reading, writing, and collaborating alongside human users. It can mean an
  agent embedded in the product (**internal**) or an open design any outside
  agent can connect to and operate in (**external**).
canonical: 'https://liquidgold.cc/channel/everyinc/explore/term/agent-native-335d4cc76755'
markdown: >-
  https://liquidgold.cc/channel/everyinc/explore/term/agent-native-335d4cc76755.md
llms: null
parent: 'https://liquidgold.cc/channel/everyinc/llms.txt'
channel: everyinc
updated: '2026-10-23T22:12:35.346Z'
language: null
children: []
---


# \*\*agent-native\*\*

Canonical: https://liquidgold.cc/channel/everyinc/explore/term/agent-native-335d4cc76755

Parent: [Channel index](https://liquidgold.cc/channel/everyinc/llms.txt)

## Definition

**agent-native** — describing software designed so AI agents can act within it — reading, writing, and collaborating alongside human users. It can mean an agent embedded in the product (**internal**) or an open design any outside agent can connect to and operate in (**external**).

In this channel, **agent-native** describes products designed so that AI agents can act within them — reading, writing, and collaborating like another user. Explore Mike Krieger's argument for agent-native design at Anthropic Labs [^dd22bafc-74bc-5611-bcc7-a7b5d85893c1@0] and Proof, Every's collaborative editor built for humans and agents together [^6fa3b152-7803-5980-8bf0-36fdcc722a1c@0].A key insight from Every's team: agent-native doesn't require an agent embedded in your product. There are two flavors: - **Internal** — an agent lives inside the product and can do anything in it. - **External** — any outside agent (like Claude) can connect and do anything in the product. Dan Shipper found the internal version weak — an embedded agent with no context isn't powerful. What works better is _your own_ context-rich Claude joining the document as a co-author [^6fa3b152-7803-5980-8bf0-36fdcc722a1c@593].Building agent-native shifts design attention to **AX (agent experience)** as much as UX — and agents themselves can tell you what confuses them [^6fa3b152-7803-5980-8bf0-36fdcc722a1c@1031].- **Agent-native** = software designed for agents to act inside it, not just humans. - Two flavors: **internal** (embedded agent) and **external** (any agent can connect) — Proof favors external. - Designing for agents means treating **agent experience (AX)** as seriously as UX.