Grain vs ChatGPT Space

Quick answer

OpenAI launched ChatGPT Space at DevDay on 2026-09-29 as the replacement for Library on Pro, Business and Enterprise plans. A space groups pages; a page is a document people and ChatGPT edit together in real time, with comments, interactive visualizations and prompt blocks. Grain separates the agent from the workspace. Your agent writes files, and Grain publishes them with public, password or invite-only sharing, comments from workspace members that the agent can read and act on, and optional Postgres, logins and scheduled jobs. Choose Space for collaborative writing inside ChatGPT. Choose Grain when you want to pick the agent, keep the work as files and read the source code.

Fit first

What is different, verifiably

Where ChatGPT Space fits

• It is already inside ChatGPT. A team that pays for Business or Enterprise gets Space without adopting, approving or paying for another tool.
• Dots. On eligible plans, an always-on agent can take a request from a page or comment and keep working in the cloud.
• Workspace administration, inherited page access and enterprise sharing controls come from the ChatGPT workspace you already manage.

Where Grain is better

• Any agent can write to it. Claude Code, Codex, Cursor and other MCP clients use the same workspace, so switching models does not strand the work.
• Artifacts are real files. You check them out, push, checkpoint, restore and pull them with the CLI, and you can diff and move them like any other folder.
• The source is open under AGPL-3.0 on GitHub. You can read how artifacts are stored and shared instead of relying on a vendor description.
• An artifact can be a running app with Postgres, logins and scheduled jobs at the same URL, not a snapshot embedded in a document.
• The free tier publishes unlimited artifacts with no card, and published artifacts stay up if you stop paying.

Moving between them

1. From ChatGPT Space to Grain: copy each page's content and download saved files. Full-fidelity page export is not documented as of 2026-09-29, so plan for comments, interactive blocks and edit history to stay behind.
2. Connect the hosted MCP server at https://api.rungrain.com/mcp from ChatGPT or Claude, or run grain init in a local folder to write MCP config for Claude Code, Codex and Cursor.
3. Ask the agent to rebuild the pages as files. Interactive blocks become a working page or app; tables that people update become a Postgres table.
4. Run grain push, grain checkpoint -m "import from ChatGPT Space" and grain share --public, or --password <pw> for a private link.
5. From Grain to ChatGPT Space: run grain pull, then paste or upload the relevant content into a new page. Grain databases, logins, scheduled jobs, share links and custom domains do not carry into Space.

Choice rule

Choose ChatGPT Space

if your team already pays for ChatGPT Business or Enterprise, wants to co-write pages in real time, and is happy for ChatGPT, Codex and dots to be the agents.

Choose Grain

if you use more than one agent, want artifacts kept as files you can export, need a running app behind the URL, or want open source code.

Use both

if the team drafts in Space, then connects Grain's hosted MCP server from ChatGPT to turn a finished page into an app or report with its own URL.

Frequently asked questions