Core setup

The home base for your agents

Core is a private and portable cloud for AI work. It gives any harness, including Codex, Claude, ChatGPT, Cursor, and others, a durable place to store memory, files, decisions, drafts, tasks, and working context. With Core, you can switch models and vendors without leaving your agents' work or reasoning behind.

Paste into your agent Choose your app

Paste this into your agent

This is the main path. Paste the request into Codex, Claude Code, Cursor, OpenCode, Hermes, or any coding agent that can help with setup. It keeps the instruction coming from you, while the guide stays as reference documentation.

Best for coding agents that can read docs, run approved commands, open browser login flows, and explain any app step that still needs you.

Please help me set up Core for this agent. Core is a home base for AI agents: it keeps memory, files, decisions, tasks, and work context synced across tools, harnesses, and models.

Use this official setup reference:
https://gradien.ai/core/docs/reference

What changes after Core is connected

Setup is meant to be a one-time step. After that, Core should fade into the background while your work compounds.

Your context survives.
Important decisions, checkpoints, and working notes do not disappear when a chat ends.
Your work stays yours.
Private work defaults to your personal Core space. Shared spaces are opt-in and permissioned.
Your agents get smarter together.
The next agent can pick up useful context from the last one instead of rebuilding the world from scratch.
Collaboration stays controlled.
Invite people to Spaces when a project, class, company, or client needs shared context.

Manual setup options

Use this when you prefer to start from a terminal instead of asking an agent to walk through setup.

Guided CLI setup

CLI install

Best when you have a terminal. The installer asks which app you use, opens Core signup or login, and applies the setup when the app supports it.

npx -y @gradien/core-agent-setup

Hosted web setup

ChatGPT web and Claude web

Open the app's connector settings and add a custom connector. Usually the URL is enough. If the app asks for more fields, use the values below.

Name: Core
Description: Personal cloud memory and files for AI agent work.
Authentication: OAuth

https://core.gradien.ai/mcp

Agent-specific guides

Each app has its own MCP settings, OAuth behavior, and restart rules. Use the guide for the app you are connecting.

How to work after setup

Keep asking for normal work. Core is useful because it lets context travel quietly with the work, not because you have to remember a new ritual.

Pick up the project where we left off.
Remember this decision for future work.
Review this against the direction I have been shaping.
Draft the next version and keep the working notes private.
Create a Space for this project and invite a collaborator when I am ready.

Explicit Core requests are still useful when you want precision: search prior decisions, save a private checkpoint, or create a Space for a team.

Privacy and collaboration

Core starts as a user product. Personal work belongs in your private cloud space by default. Collaboration appears through Spaces you create or join.

Agents can search material you are allowed to see.
Agents should only change shared or team material when you explicitly ask.
Space owners can invite people and choose read or write permissions.
A user can belong to many Spaces, such as a company, class, client project, or team.

Reference

Use these when setting up custom harnesses or debugging a connector.

Core setup guideRendered setup guide for agents and humans. TroubleshootingRecovery steps for OAuth, reload, and connector issues. Full agent-readable contextComplete public Core product and setup context. OAuth protected resource metadataDiscovery metadata for remote MCP clients.

What Is Core Plain-language explanation of Core. Why Core Matters Why personal agent memory and controlled shared context matter. How To Use Core Daily operating loop for external agents. Core FAQ Frequently asked product and setup questions.