Grain for Claude Code

# Download the Mac app from rungrain.com, then:
grain login
cd path/to/project
grain init
# Ask Claude Code to create the artifact in this folder.
grain push
grain checkpoint -m "first Claude Code artifact"
grain share --public
# https://<slug>.share.rungrain.com

What Claude Code does, what Grain does

Setup

1. Download the Mac app from rungrain.com, open it and complete sign-in. The download dialog asks for an early-access phrase and provides a request-access form if you do not have one.
2. Open a terminal, run grain login, and finish the authentication flow before creating a workspace. Keep the desktop app available while you set up the folder.
3. Change into the project directory and run grain init. For Claude Code, this writes the workspace connection to .mcp.json. If that file already carries other MCP servers, inspect the result and keep the existing entries alongside Grain.
4. Start Claude Code in the same project. Give it a bounded artifact request, including the intended reader, required inputs, and whether the result is a static page or an app that needs Postgres, login or a scheduled job.
5. Ask the agent to use the connected Grain tools when it needs to open, push, checkpoint or share work. You can also run the CLI sequence yourself, which makes every publishing step visible in terminal history.

First artifact

1. Use a concrete prompt: Build a PR walkthrough for the current branch. Create a page with a plain-language summary, changed-file table, test checklist and known limits. Keep it useful to a reviewer who has not read the chat. This tells Claude Code what to make and who will open it.
2. Review the generated files locally. Ask the agent to remove secrets, local paths and temporary logs. For an app, decide explicitly whether it needs Postgres, login or a schedule. Do not add those pieces just because they are available.
3. Run grain push. This sends the workspace files after the agent has written them. Then run grain checkpoint -m "first Claude Code artifact" so the published state has a named point you can return to.
4. Run grain share --public for a public link, or grain share --password <pw> when the reader should supply a password. Invited-person sharing is the third access mode when the audience must be named.
5. Open the returned https://<slug>.share.rungrain.com address outside your signed-in session. Check the page, its access boundary and any data-backed interaction. Send that URL rather than a screenshot or localhost recording.
6. When feedback arrives as comments, have Claude Code read it and make the next change. Push and checkpoint again so the same workspace accumulates revisions instead of creating a disconnected artifact for every review round.

What you pay

Your Claude plan remains separate and under your control. Grain does not sell or mark up agent tokens. Grain's free tier includes the Mac app, CLI, unlimited published artifacts, ten sandboxes that sleep when idle, one always-on workspace and a self-hostable core, with no card required. Pro is $65 per user per month and adds more always-on workspaces, team seats, shared spaces and custom domains. Enterprise is available by contact for deployment in the customer's cloud with SSO, audit and policy controls and direct support. Paying Grain does not change which Claude plan or model you use.

Limits and known gaps

• Grain is not Claude Code and does not generate the artifact. If the prompt, code or tests are wrong, publishing preserves the wrong result until you change it.
• Grain is not an IDE. Keep your editor and Claude Code terminal workflow for human code review, debugging and source control.
• The desktop app is Mac first. Windows has a waitlist, while the self-hostable core runs on Linux.
• A public share is reachable by anyone with the link. Use a password or invited people when the artifact contains material that should not be public.
• The .mcp.json connection is specific to Claude Code. Do not copy a Cursor or Codex config path into this setup and assume the client will read it.
• Grain is not a general deploy platform. For a production service that needs its own CI pipeline and deployment controls, use a deployment platform instead.

Frequently asked questions