Endue Docs

Endue is a platform for AI agents that do real work. You give an agent an identity and a purpose, connect it to the tools you already use, and it runs — in a conversation, on a schedule, or triggered by an event — while you keep control of every action that leaves your account.

Get started

Endue runs on several surfaces. They all talk to the same agents, conversations, and history, so you can start on one and continue on another. Pick where you want to begin.

Web

The full Endue workspace in your browser. Build agents, chat with them, connect tools, schedule routines, and review everything they produce — no install.

Open Endue and sign in, then follow the Quickstart to build your first agent.

This is the recommended place to start, and the surface the rest of these docs show by default.

Desktop

A native desktop app for macOS, Windows, and Linux. Same workspace as the web, plus the ability to host agent runs on your own machine so they can reach local files and tools.

Setup docs in progress

Desktop install and pairing instructions are being written. See Surfaces for what the desktop app does today.

Mobile

iOS and Android apps for checking in on agents that are already running: read what they produced, answer a question an agent asked, and approve or reject an action it wants to take.

Setup docs in progress

Mobile install instructions are being written. See Surfaces.

CLI

A single endue binary that is both a local runtime and a command-line client — endue chat for an interactive session, endue run for one-shot tasks you can pipe into other tools.

Setup docs in progress

The CLI is not yet generally available. See Surfaces for the commands it exposes.

What you can do

Give an agent an identity, not just a prompt

An Endue agent is a durable thing, not a chat window. It has a name and handle, a profile, a character, and a system prompt kept under revision history — so you can change how it behaves and roll back when a change makes it worse.

Agents you build are addressable by handle, and the same agent shows up wherever you work with it.

→ Build your agent

Connect the tools you already use

Connect an agent to more than thirty services — Gmail and Google Workspace, Slack, Notion, Linear, GitLab, Shopify, Figma, Datadog and the rest. Once connected, the agent can search, read, create, and send within the accounts you authorized — nothing more.

Each connection is scoped to your account, bound to the agents you choose, and can be revoked at any time.

→ Available connectors

Teach it skills instead of rewriting prompts

A skill packages a repeatable job — a checklist, a procedure, a set of commands — so you can bind it to an agent and reuse it. An agent can also ask for a skill it is not yet allowed to use, and you decide in the conversation whether to grant it.

Let it remember what matters

Agents can store what they learn about you and your work, and recall it in later conversations. Memory is explicit and inspectable: you can see what an agent remembered and remove anything that should not have been kept.

Put recurring work on a schedule

Turn a task into a routine that runs on a schedule — a morning digest, a weekly report, an overnight check. Routines run whether or not you are in the app, and what they produce lands in your workspace waiting for you.

Get artifacts, not walls of text

When an agent produces something worth keeping — a report, a document, a diagram — it becomes an artifact with its own page and a shareable link. You can hand that link to someone outside your workspace without giving them access to anything else.

→ Artifacts

Stay in control while it runs

You are not locked out once a run starts. Steer it mid-flight with a correction, and approve or reject any action that sends something out of your account or destroys data. Nothing outbound happens without your say-so.

→ Core concepts: Run

Organize work into projects and spaces

Group related conversations, artifacts, and routines into a project so an agent has the right context. Spaces let a team share agents and the work they produce.

Find the right page

I want to…	Start here
Build my first agent and talk to it	Quickstart
Understand what an agent, run, or artifact actually is	Core concepts
Know why an agent did what it did	How Endue works
Write a system prompt that actually works	System prompt and revisions
Get better results from an agent I already built	Testing and iterating
Connect Gmail, Slack, Notion, or something else	Available connectors
Put a recurring job on a schedule	Routines
Understand what an agent can and cannot reach	Security and permissions
Work out why something went wrong	Troubleshooting
Know which app to install	Surfaces
Look up a term I saw in the product	Glossary

Reading these docs with an agent

Every page is available as plain Markdown, and the whole site is indexed at /llms.txt and /llms-full.txt. Point an agent at either one and it can answer questions about Endue from the current docs.

Next steps

QuickstartBuild an agent, connect a tool, and get your first real result.

Core conceptsAgents, conversations, runs, skills, connectors, memory, artifacts.

How Endue worksThe loop behind every run, what an agent can see, and why runs pause.

Build your agentIdentity, system prompt, model — everything that decides how an agent behaves.

Connect your toolsMore than thirty services, and the binding model that keeps access narrow.

Getting help

• Search these docs — press Ctrl/⌘ + K from any page.
• Look up a term — the Glossary defines everything you can see in the product.
• Email us — [email protected].