Booklet MCP Server

Booklet

Publish clean, readable pages from Markdown — by hand, by script, or by your AI assistant.

Paste Markdown into the editor, preview it live, and share a polished read-only URL in one click — or skip the editor and publish the same way from a CI pipeline, a script, or an AI assistant like Claude. Sign in for pages that never expire, version history, analytics, custom slugs, password protection, collections, and a full REST API.

What makes Booklet more than an editor is the surface around it: a REST API, an npm-published CLI, a GitHub Action for publishing docs in CI, and a standalone MCP server so AI assistants like Claude can publish and manage pages directly. Markdown-to-shareable-page tools are common; shipping the same functionality as an API, a CLI, a CI action, and an MCP server on top of it is the part that isn't.

Live: booklet.ashwinsathian.com · API docs: /api-docs · MCP setup: /mcp-setup

---

Quick start

npm install -g booklet-cli
booklet login                            # opens your browser to authorize
booklet publish README.md --open         # publish this file, open it in your browser

That's it: you get back a permanent, shareable URL. No account needed to try the editor itself; sign in only when you want pages that never expire, an API key, or the CLI.

---

Features

• Editor: live preview (120 ms debounce), unlimited local drafts with autosave, import/export

• Share pages: clean read-only URLs, table of contents, reading time, dark/light mode

• Embeds: <iframe> embed codes for any page via /p/:id/embed

• Export: PDF, Markdown, HTML fragment

• LaTeX / KaTeX: inline $...$ and display $$...$$ math blocks

• Mermaid diagrams: fenced code blocks with mermaid language tag

• Version history: every publish is snapshotted; browse and restore past versions

• Analytics: per-page view counts, scroll depth, referrers

• Collections: group pages into a named collection with a shared URL

• Password protection: require a password to view any page

• Custom slugs: set a human-readable URL like /p/my-release-notes

• Team Spaces: invite collaborators, publish to shared /t/:slug spaces

• Webhooks: HTTP callbacks on page.published and page.updated events

• REST API: publish, update, list, and delete pages programmatically

• CLI: publish Markdown from your terminal (npx booklet-cli)

• GitHub Action: publish docs in CI via AshwinSathian/publish-to-booklet

• MCP server: expose the API to AI assistants (Claude, Cursor, etc.) via the Model Context Protocol

• Frontmatter: YAML frontmatter sets title, slug, visibility, tags, author, date

---

CLI

npm install -g booklet-cli

booklet login                          # save your API key
booklet publish README.md              # publish a file
booklet publish README.md --watch      # watch + auto-republish on save
booklet publish - < NOTES.md           # from stdin
booklet pages list                     # list your pages

See packages/cli/README.md for full docs (all flags, CI/non-interactive auth via --key or BOOKLET_API_KEY, pages open, etc.).

---

REST API

All endpoints are under /api/v1/ and authenticated with Authorization: Bearer <bklt_...>.

Publish example:

curl -X POST https://booklet-api.ashwinsathian.com/api/v1/publish \
  -H "Authorization: Bearer bklt_..." \
  -H "Content-Type: application/json" \
  -d '{"raw": "# Hello\n\nThis is my page."}'

booklet-api.ashwinsathian.com is a dedicated hostname for the API surface (same app/process as the main site, just scoped; see docs/OPERATIONS.md). booklet.ashwinsathian.com serves /api/v1/* too, so either works.

Full endpoint reference with request/response shapes: booklet.ashwinsathian.com/api-docs.

---

MCP Server

A plain Node process (mcp-server/) that exposes Booklet's API to AI assistants supporting the Model Context Protocol, run under PM2 alongside the main app, not a Cloudflare Worker (that was the original design, changed when the rest of the app moved off Cloudflare Workers; see docs/OPERATIONS.md).

Endpoint: https://booklet-mcp.ashwinsathian.com/mcp Auth: Authorization: Bearer <bklt_...> header (same API keys as the REST API) Tools: publish_page, update_page, get_page, list_pages, delete_page Resources: published pages are also exposed as browsable/readable MCP resources (booklet://pages/:id) Prompts: five pre-built document templates the assistant can fill in and publish directly — incident_report, adr (Architecture Decision Record), release_notes, rfc, and runbook

Point any MCP-compatible client at the endpoint above with your API key in the Authorization header. booklet.ashwinsathian.com/mcp-setup has copy-paste config for Claude Desktop, Claude.ai, Cursor, Windsurf, VS Code, and Zed.

To run the server itself locally:

cd mcp-server && npm run dev

---

Tech stack

---

Local development

Prerequisites

• Node.js 20+
• MongoDB connection string (a local mongod, or any self-hosted/managed instance)

Install & run

npm install
npm run dev        # Next.js dev server at http://localhost:3000

Environment variables

Create .env.local:

NEXT_PUBLIC_SITE_URL=http://localhost:3000
NEXT_PUBLIC_GA_MEASUREMENT_ID=G-XXXXXXXXXX
MONGODB_URI=mongodb://localhost:27017/booklet

# Required: dedicated secret that signs/verifies team-invite JWT tokens.
# Must be its own random value; there is no fallback, and invite creation
# and joining fail closed if this is unset. Generate with: openssl rand -base64 32
INVITE_JWT_SECRET=<random-secret>

See .env.example for the full list of required secrets (session auth, API keys, page-unlock tokens, etc.). Each documents its own generation command and fail-closed behavior.

Deploy

npm run deploy     # rebuilds and restarts the PM2-managed app + MCP server (scripts/redeploy.sh)

---

Project structure

src/
  app/
    app/            # Editor (client)
    p/[id]/         # Share page + embed
    my-pages/       # Dashboard: pages, API keys, webhooks, collections
    api/v1/         # REST API
    explore/        # Public page directory
    templates/      # Template landing pages
  components/
    blocks/         # BlockRenderer + InlineRenderer (custom AST renderer)
    share/          # TOC, export, embed, reading progress, analytics beacon
    ui/             # Design system components
  lib/
    blocks.ts       # Block/Inline type definitions
    parse.ts        # Markdown → Block[] (unified pipeline)
    db/             # MongoDB helpers
    storage.ts      # Document content storage (MongoDB)
    quota.ts        # Feature flags
    frontmatter.ts  # YAML frontmatter parser (js-yaml)
packages/           # npm workspaces; one root lockfile covers all of these
  shared/           # booklet-api-client: shared /api/v1 schemas + client
  cli/              # booklet-cli npm package
  vscode/           # VS Code extension: publish from editor
mcp-server/         # MCP server (plain Node process, run under PM2)
.github/
  workflows/        # ci.yml, publish-cli.yml, publish-shared.yml, publish-vscode.yml
  examples/         # publish-to-booklet.yml, use in your own repo

---

GitHub Actions

CI

Every push/PR to main runs lint, typecheck (root app + each workspace package), a production build, and the unit test suite against a real MongoDB service container. See .github/workflows/ci.yml.

Auto-publish to npm

Push to main with a bumped version in packages/cli/package.json or packages/shared/package.json → automatically publishes booklet-cli or booklet-api-client to npm.

Required secret: NPM_TOKEN (Granular Access Token with publish + 2FA bypass).

Publish docs to Booklet from your repo

Two ways: the AshwinSathian/publish-to-booklet GitHub Action, or booklet-cli via npx. See .github/examples/publish-to-booklet.yml for both — copy it into your own repo's .github/workflows/, add a BOOKLET_API_KEY secret, and it publishes on every release.