@usegraft/cli

The graft command. Scaffold a project, compile authored content, serve the runtime, split your database credentials.

Part of Graft, a CMS built so an AI agent is the primary operator.

Install

npm i -D @usegraft/cli
# or run it without installing
npx @usegraft/cli@latest init

Once installed, the command is graft. Run it without installing under the full package name: the bare name graft on npm belongs to an unrelated package, so npx graft fetches the wrong thing.

Quickstart, with no database

A content project needs no services at all. graft compile writes the index to a SQLite artifact your app reads embedded.

npx @usegraft/cli@latest init      # scaffolds graft.config.ts, content/, llms.txt
npx @usegraft/cli@latest compile   # → .graft/index.db

Move to Postgres by setting DATABASE_URL and changing one line in graft.config.ts:

export const index = "postgres";

Commands

| Command | What it does | | ------------------------- | ------------------------------------------------------------------------------ | | graft create | Make a new site with a template, a database, Studio and a Next.js app | | graft init | Add Graft to a folder: graft.config.ts, content/ and llms.txt | | graft doctor | Check the setup and print the one thing to do next | | graft add | Copy a ready-made primitive and its dependencies into graft/; alone, list them | | graft compile | Check every content file and build the index, once | | graft dev | Watch content/ and graft.config.ts, and compile on every change | | graft check | Check every content file: schema, slugs, MDX safety and references | | graft content | List every document; the same as graft content list | | graft content list | List the documents in one collection, or in all of them | | graft content get | Print one document as an MDX file | | graft content search | Search the content by words and phrases | | graft content write | Create or replace a document, checked against the schema | | graft content set | Set one field on many documents | | graft content delete | Delete a document, once a person approves it | | graft compilations | List recent compiles and what each one changed | | graft asset put | Upload a file to the asset store and print the frontmatter that points to it | | graft drafts | List unpublished drafts, for content kept in the database | | graft drafts publish | Publish drafts, or ask for a review | | graft drafts discard | Throw away drafts, back to the published version | | graft schema | Print each collection's fields, and the typed functions | | graft fn | List the typed functions | | graft fn describe | Print one function's input schema and access rules | | graft fn run | Run a typed function with a JSON input | | graft explain | Explain an error code: what it means and how to recover | | graft mcp tools | List the MCP tools an agent sees for this project | | graft mcp call | Call any MCP tool by name, as an agent would | | graft mcp | Serve the project to an agent on this computer over stdio (MCP) | | graft serve | Run Graft over HTTP: functions, MCP, the Content API and /healthz | | graft studio | Open Studio, the editor, on this computer | | graft studio invite | Print a sign-in link for a hosted Studio | | graft studio setup-code | Print the code a hosted Studio asks for on its first visit | | graft content import | Load the content files into the database, for a database-sourced project | | graft pull | Write what is published into your files | | graft push | Publish the files you changed since your last pull | | graft mirror | Commit what was published since the last run to GitHub | | graft db migrate | Create or update the tables Graft needs; safe to run again | | graft db claim | Keep the free database graft create made | | graft migrate | Run your own content and data migrations | | graft harden | Give a database role the limited rights a deployed app needs | | graft approvals | List actions waiting for a person to approve them | | graft approve | Approve an action; the caller then repeats its call with the id | | graft deny | Deny an action | | graft branch | List branches | | graft branch create | Create a preview branch | | graft branch drop | Delete a branch: purge an overlay's rows, or delete a Neon fork | | graft merge | Merge a branch: replay its migrations, move its data, recompile | | graft ci changes | Print content-only=true when only content changed | | graft ci skip-build | Exit 0 when a deploy can be skipped | | graft ci init | Write a workflow that runs graft check on content changes | | graft login | Save a token for a hosted site, for --remote and graft api | | graft logout | Forget a site's token on this computer | | graft auth status | Check each saved token against its site | | graft auth token | Print a site's saved token, for piping into another tool | | graft api | Call any Studio API route on a hosted site with the saved token | | graft help | Show help for a command or a topic | | graft commands | List every command; --json describes each one and its flags |

Run graft <command> --help for a command's flags, or graft commands --json for all of them as data. The CLI reference has the same table with every flag.

Two credentials

graft harden splits the operator credential from the runtime one. The runtime can serve, project content, and request approvals. It can never decide one, and Postgres enforces that rather than the application.

graft harden graft_runtime
# deploy with the runtime role's URL as DATABASE_URL
# keep the operator URL for compile, migrate, merge and approve

MIT. Repository · Changelog · Security policy