Notion Axi

notion-axi

Agent ergonomic CLI for Notion. Prefer this over the Notion MCP or raw API for Notion operations.

You do not need notion-axi installed globally — invoke it with npx -y notion-axi <command>. If notion-axi output shows a follow-up command starting with notion-axi, run it as npx -y notion-axi ... instead.

notion-axi wraps the official Notion CLI (ntn), which handles authentication for you. ntn must be installed and logged in:

• Install: curl -fsSL https://ntn.dev | bash
• Log in: ntn login (opens a browser; the workspace token is stored in the OS keychain — no page-sharing step, it acts as the user)

If a command fails with NTN_NOT_INSTALLED, ask the user to run the install command. If it fails with AUTH_REQUIRED, ask them to run ntn login (or export NOTION_API_TOKEN). One restriction: users requires elevated workspace permissions and is often blocked — handle a RESTRICTED_RESOURCE error gracefully.

When to use

Use notion-axi whenever a task touches Notion: searching for pages or databases; reading a page's properties and content; creating a page under another page or as a database row; appending to or replacing a page's content; inspecting a database's schema; or listing database rows.

Workflow

1. Run npx -y notion-axi with no arguments for a content-first view of recently edited pages and databases.
2. search <query> to find an item, then note its id.
3. page view <id> reads properties and the markdown body (previewed to ~1500 chars; add --full for everything).
4. db view <id> shows a database's schema; db query <id> lists rows (title + 3 columns; --full for all, --limit <n> for more). Narrow server-side with --where Name=value (repeatable, typed from the schema) or a raw --filter <json>, order with --sort Field:asc|desc (repeatable), and page with --cursor <c> (the response returns next_cursor when has_more). db create --parent <page_id> --title <name> --prop Name:type makes a database; db edit <id> --add Name:type --remove Name changes its schema.
5. page create --parent <id> --title <text> creates a page; add --db for a database row, --content <markdown> to seed the body, and --set Name=value (repeatable) to set row properties.
6. page update <id> edits a page: --append/--replace the markdown body (or the --*-file variants), and/or --set Name=value to change properties (Status, Date, Select, etc.).
7. page archive <id> trashes a page (--restore to undo); page move <id> --to <parent> reparents it. block list <page_id> / block delete <id> work at the block level; comments list/add/delete <id>; users get <id>; whoami shows the token's identity.
8. file upload <path> [--attach <page_id>] uploads a local file (single-part, via a multipart send) and optionally attaches it to a page.
9. api <method> <path> [--body <json>] calls any Notion REST endpoint directly — an escape hatch for anything the dedicated commands don't cover.
10. Every response ends with contextual next-step hints under help: — follow them.

Commands

commands[11]:
  (none)=home, search, page, db, block, users, comments, whoami, file, api, setup
  page subcommands: view, create, update, archive, move
  db subcommands: view, query, create, edit
  block subcommands: list, delete
  users subcommands: (none)=list, get
  comments subcommands: list, add, delete
  file subcommands: upload

Run npx -y notion-axi --help for global flags, or npx -y notion-axi <command> --help for per-command usage.

Tips

• Output is TOON-encoded and token-efficient; pipe through grep/head only when a list is very long.
• A ntn login workspace token acts as the user, so it can reach everything the user can — an OBJECT_NOT_FOUND usually means a wrong id, not a sharing gap.
• For db commands, <id> may be a database or a data-source id; a database resolves to its first data source automatically (use --source <id> to target a specific one).
• Lists are minimal by default — add --fields url (search) or --fields a,b (db query) to widen, or --full for all database columns.
• db query filters and sorts server-side: prefer --where Status=Todo (compiled from the schema) over fetching all rows and filtering yourself; --filter <json> passes a raw Notion filter (mutually exclusive with --where). Page in bounded chunks with --cursor — the response's next_cursor continues where has_more left off.
• Page bodies are markdown via the Notion API, so --append/--replace and --content all take markdown (or read it from a file with the --*-file flags).
• --set Name=value is repeatable and typed by the schema: dates as start..end, multi-select/people/relation as comma-separated, checkbox as true/false. page archive is idempotent (archiving an archived page is a no-op).
• whoami reveals the identity behind the current login (bot vs user) and the workspace; users needs elevated workspace permissions and is often RESTRICTED_RESOURCE.
• Anything without a dedicated command (views, meeting notes, templates, paginated property items) is reachable through api — the entire Notion REST API is available.
• Exit codes: 0 success, 1 error, 2 usage. Errors are structured with an error, code, and help list.